diff --git a/is/is_IS/bui/medium/ALIASES b/is/is_IS/bui/medium/ALIASES new file mode 100644 index 0000000000000000000000000000000000000000..9337bb64ab563648fb2e213a59585a25db1f983d --- /dev/null +++ b/is/is_IS/bui/medium/ALIASES @@ -0,0 +1 @@ +is-bui-medium diff --git a/is/is_IS/bui/medium/MODEL_CARD b/is/is_IS/bui/medium/MODEL_CARD new file mode 100644 index 0000000000000000000000000000000000000000..52def23e17df8fc801fb094f1c9b4cae87bdae11 --- /dev/null +++ b/is/is_IS/bui/medium/MODEL_CARD @@ -0,0 +1,15 @@ +# Model card for bui (medium) + +* Language: is_IS (Icelandic, Iceland) +* Speakers: 1 (Búi) +* Quality: medium +* Samplerate: 22,050Hz + +## Dataset + +* URL: http://hdl.handle.net/20.500.12537/104 +* License: See URL + +## Training + +Trained from scratch. diff --git a/is/is_IS/bui/medium/is_IS-bui-medium.onnx b/is/is_IS/bui/medium/is_IS-bui-medium.onnx new file mode 100644 index 0000000000000000000000000000000000000000..d8f3d2dc78bcee3c3a22fd78fad29bdb2b35e6d0 --- /dev/null +++ b/is/is_IS/bui/medium/is_IS-bui-medium.onnx @@ -0,0 +1,3 @@ +version https://git-lfs.github.com/spec/v1 +oid sha256:3a645b2d2850e4098f01f3765cece931836c03741e01a5cc514d09d39d37c05c +size 76495465 diff --git a/is/is_IS/bui/medium/is_IS-bui-medium.onnx.json b/is/is_IS/bui/medium/is_IS-bui-medium.onnx.json new file mode 100644 index 0000000000000000000000000000000000000000..126e455c143b2172e9b654ded201cdbc7be59d85 --- /dev/null +++ b/is/is_IS/bui/medium/is_IS-bui-medium.onnx.json @@ -0,0 +1,420 @@ +{ + "audio": { + "sample_rate": 22050, + "quality": "medium" + }, + "espeak": { + "voice": "is" + }, + "inference": { + "noise_scale": 0.667, + "length_scale": 1, + "noise_w": 0.8 + }, + "phoneme_map": {}, + "phoneme_id_map": { + "_": [ + 0 + ], + "^": [ + 1 + ], + "$": [ + 2 + ], + " ": [ + 3 + ], + "!": [ + 4 + ], + "'": [ + 5 + ], + "(": [ + 6 + ], + ")": [ + 7 + ], + ",": [ + 8 + ], + "-": [ + 9 + ], + ".": [ + 10 + ], + ":": [ + 11 + ], + ";": [ + 12 + ], + "?": [ + 13 + ], + "a": [ + 14 + ], + "b": [ + 15 + ], + "c": [ + 16 + ], + "d": [ + 17 + ], + "e": [ + 18 + ], + "f": [ + 19 + ], + "h": [ + 20 + ], + "i": [ + 21 + ], + "j": [ + 22 + ], + "k": [ + 23 + ], + "l": [ + 24 + ], + "m": [ + 25 + ], + "n": [ + 26 + ], + "o": [ + 27 + ], + "p": [ + 28 + ], + "q": [ + 29 + ], + "r": [ + 30 + ], + "s": [ + 31 + ], + "t": [ + 32 + ], + "u": [ + 33 + ], + "v": [ + 34 + ], + "w": [ + 35 + ], + "x": [ + 36 + ], + "y": [ + 37 + ], + "z": [ + 38 + ], + "æ": [ + 39 + ], + "ç": [ + 40 + ], + "ð": [ + 41 + ], + "ø": [ + 42 + ], + "ħ": [ + 43 + ], + "ŋ": [ + 44 + ], + "œ": [ + 45 + ], + "ǀ": [ + 46 + ], + "ǁ": [ + 47 + ], + "ǂ": [ + 48 + ], + "ǃ": [ + 49 + ], + "ɐ": [ + 50 + ], + "ɑ": [ + 51 + ], + "ɒ": [ + 52 + ], + "ɓ": [ + 53 + ], + "ɔ": [ + 54 + ], + "ɕ": [ + 55 + ], + "ɖ": [ + 56 + ], + "ɗ": [ + 57 + ], + "ɘ": [ + 58 + ], + "ə": [ + 59 + ], + "ɚ": [ + 60 + ], + "ɛ": [ + 61 + ], + "ɜ": [ + 62 + ], + "ɞ": [ + 63 + ], + "ɟ": [ + 64 + ], + "ɠ": [ + 65 + ], + "ɡ": [ + 66 + ], + "ɢ": [ + 67 + ], + "ɣ": [ + 68 + ], + "ɤ": [ + 69 + ], + "ɥ": [ + 70 + ], + "ɦ": [ + 71 + ], + "ɧ": [ + 72 + ], + "ɨ": [ + 73 + ], + "ɪ": [ + 74 + ], + "ɫ": [ + 75 + ], + "ɬ": [ + 76 + ], + "ɭ": [ + 77 + ], + "ɮ": [ + 78 + ], + "ɯ": [ + 79 + ], + "ɰ": [ + 80 + ], + "ɱ": [ + 81 + ], + "ɲ": [ + 82 + ], + "ɳ": [ + 83 + ], + "ɴ": [ + 84 + ], + "ɵ": [ + 85 + ], + "ɶ": [ + 86 + ], + "ɸ": [ + 87 + ], + "ɹ": [ + 88 + ], + "ɺ": [ + 89 + ], + "ɻ": [ + 90 + ], + "ɽ": [ + 91 + ], + "ɾ": [ + 92 + ], + "ʀ": [ + 93 + ], + "ʁ": [ + 94 + ], + "ʂ": [ + 95 + ], + "ʃ": [ + 96 + ], + "ʄ": [ + 97 + ], + "ʈ": [ + 98 + ], + "ʉ": [ + 99 + ], + "ʊ": [ + 100 + ], + "ʋ": [ + 101 + ], + "ʌ": [ + 102 + ], + "ʍ": [ + 103 + ], + "ʎ": [ + 104 + ], + "ʏ": [ + 105 + ], + "ʐ": [ + 106 + ], + "ʑ": [ + 107 + ], + "ʒ": [ + 108 + ], + "ʔ": [ + 109 + ], + "ʕ": [ + 110 + ], + "ʘ": [ + 111 + ], + "ʙ": [ + 112 + ], + "ʛ": [ + 113 + ], + "ʜ": [ + 114 + ], + "ʝ": [ + 115 + ], + "ʟ": [ + 116 + ], + "ʡ": [ + 117 + ], + "ʢ": [ + 118 + ], + "ʲ": [ + 119 + ], + "ˈ": [ + 120 + ], + "ˌ": [ + 121 + ], + "ː": [ + 122 + ], + "ˑ": [ + 123 + ], + "˞": [ + 124 + ], + "β": [ + 125 + ], + "θ": [ + 126 + ], + "χ": [ + 127 + ], + "ᵻ": [ + 128 + ], + "ⱱ": [ + 129 + ] + }, + "num_symbols": 130, + "num_speakers": 1, + "speaker_id_map": {}, + "piper_version": "0.2.0", + "language": { + "code": "is_IS", + "family": "is", + "region": "IS", + "name_native": "íslenska", + "name_english": "Icelandic", + "country_english": "Iceland" + }, + "dataset": "bui" +} \ No newline at end of file diff --git a/is/is_IS/bui/medium/samples/speaker_0.mp3 b/is/is_IS/bui/medium/samples/speaker_0.mp3 new file mode 100644 index 0000000000000000000000000000000000000000..304e4a486e65fe7f98ca31aec2806b7126620773 Binary files /dev/null and b/is/is_IS/bui/medium/samples/speaker_0.mp3 differ diff --git a/is/is_IS/salka/medium/ALIASES b/is/is_IS/salka/medium/ALIASES new file mode 100644 index 0000000000000000000000000000000000000000..c1563ac7e8ce46f0867da42920fc7f6467cf1151 --- /dev/null +++ b/is/is_IS/salka/medium/ALIASES @@ -0,0 +1 @@ +is-salka-medium diff --git a/is/is_IS/salka/medium/MODEL_CARD b/is/is_IS/salka/medium/MODEL_CARD new file mode 100644 index 0000000000000000000000000000000000000000..c4deaa6c0c78772c862624d124bf08bf131537e1 --- /dev/null +++ b/is/is_IS/salka/medium/MODEL_CARD @@ -0,0 +1,15 @@ +# Model card for salka (medium) + +* Language: is_IS (Icelandic, Iceland) +* Speakers: 1 +* Quality: medium +* Samplerate: 22,050Hz + +## Dataset + +* URL: http://hdl.handle.net/20.500.12537/104 +* License: See URL + +## Training + +Trained from scratch. diff --git a/is/is_IS/salka/medium/is_IS-salka-medium.onnx b/is/is_IS/salka/medium/is_IS-salka-medium.onnx new file mode 100644 index 0000000000000000000000000000000000000000..cb3f8e85c00d8b80926287f2781ad22e44170cfe --- /dev/null +++ b/is/is_IS/salka/medium/is_IS-salka-medium.onnx @@ -0,0 +1,3 @@ +version https://git-lfs.github.com/spec/v1 +oid sha256:392a8e033772de8f7d4afa3258580f0fd509678b55014c3dc1a0d3e231475d0e +size 76495465 diff --git a/is/is_IS/salka/medium/is_IS-salka-medium.onnx.json b/is/is_IS/salka/medium/is_IS-salka-medium.onnx.json new file mode 100644 index 0000000000000000000000000000000000000000..1f2234d27245fb3478435376df96eec8837b6f37 --- /dev/null +++ b/is/is_IS/salka/medium/is_IS-salka-medium.onnx.json @@ -0,0 +1,420 @@ +{ + "audio": { + "sample_rate": 22050, + "quality": "medium" + }, + "espeak": { + "voice": "is" + }, + "inference": { + "noise_scale": 0.667, + "length_scale": 1, + "noise_w": 0.8 + }, + "phoneme_map": {}, + "phoneme_id_map": { + "_": [ + 0 + ], + "^": [ + 1 + ], + "$": [ + 2 + ], + " ": [ + 3 + ], + "!": [ + 4 + ], + "'": [ + 5 + ], + "(": [ + 6 + ], + ")": [ + 7 + ], + ",": [ + 8 + ], + "-": [ + 9 + ], + ".": [ + 10 + ], + ":": [ + 11 + ], + ";": [ + 12 + ], + "?": [ + 13 + ], + "a": [ + 14 + ], + "b": [ + 15 + ], + "c": [ + 16 + ], + "d": [ + 17 + ], + "e": [ + 18 + ], + "f": [ + 19 + ], + "h": [ + 20 + ], + "i": [ + 21 + ], + "j": [ + 22 + ], + "k": [ + 23 + ], + "l": [ + 24 + ], + "m": [ + 25 + ], + "n": [ + 26 + ], + "o": [ + 27 + ], + "p": [ + 28 + ], + "q": [ + 29 + ], + "r": [ + 30 + ], + "s": [ + 31 + ], + "t": [ + 32 + ], + "u": [ + 33 + ], + "v": [ + 34 + ], + "w": [ + 35 + ], + "x": [ + 36 + ], + "y": [ + 37 + ], + "z": [ + 38 + ], + "æ": [ + 39 + ], + "ç": [ + 40 + ], + "ð": [ + 41 + ], + "ø": [ + 42 + ], + "ħ": [ + 43 + ], + "ŋ": [ + 44 + ], + "œ": [ + 45 + ], + "ǀ": [ + 46 + ], + "ǁ": [ + 47 + ], + "ǂ": [ + 48 + ], + "ǃ": [ + 49 + ], + "ɐ": [ + 50 + ], + "ɑ": [ + 51 + ], + "ɒ": [ + 52 + ], + "ɓ": [ + 53 + ], + "ɔ": [ + 54 + ], + "ɕ": [ + 55 + ], + "ɖ": [ + 56 + ], + "ɗ": [ + 57 + ], + "ɘ": [ + 58 + ], + "ə": [ + 59 + ], + "ɚ": [ + 60 + ], + "ɛ": [ + 61 + ], + "ɜ": [ + 62 + ], + "ɞ": [ + 63 + ], + "ɟ": [ + 64 + ], + "ɠ": [ + 65 + ], + "ɡ": [ + 66 + ], + "ɢ": [ + 67 + ], + "ɣ": [ + 68 + ], + "ɤ": [ + 69 + ], + "ɥ": [ + 70 + ], + "ɦ": [ + 71 + ], + "ɧ": [ + 72 + ], + "ɨ": [ + 73 + ], + "ɪ": [ + 74 + ], + "ɫ": [ + 75 + ], + "ɬ": [ + 76 + ], + "ɭ": [ + 77 + ], + "ɮ": [ + 78 + ], + "ɯ": [ + 79 + ], + "ɰ": [ + 80 + ], + "ɱ": [ + 81 + ], + "ɲ": [ + 82 + ], + "ɳ": [ + 83 + ], + "ɴ": [ + 84 + ], + "ɵ": [ + 85 + ], + "ɶ": [ + 86 + ], + "ɸ": [ + 87 + ], + "ɹ": [ + 88 + ], + "ɺ": [ + 89 + ], + "ɻ": [ + 90 + ], + "ɽ": [ + 91 + ], + "ɾ": [ + 92 + ], + "ʀ": [ + 93 + ], + "ʁ": [ + 94 + ], + "ʂ": [ + 95 + ], + "ʃ": [ + 96 + ], + "ʄ": [ + 97 + ], + "ʈ": [ + 98 + ], + "ʉ": [ + 99 + ], + "ʊ": [ + 100 + ], + "ʋ": [ + 101 + ], + "ʌ": [ + 102 + ], + "ʍ": [ + 103 + ], + "ʎ": [ + 104 + ], + "ʏ": [ + 105 + ], + "ʐ": [ + 106 + ], + "ʑ": [ + 107 + ], + "ʒ": [ + 108 + ], + "ʔ": [ + 109 + ], + "ʕ": [ + 110 + ], + "ʘ": [ + 111 + ], + "ʙ": [ + 112 + ], + "ʛ": [ + 113 + ], + "ʜ": [ + 114 + ], + "ʝ": [ + 115 + ], + "ʟ": [ + 116 + ], + "ʡ": [ + 117 + ], + "ʢ": [ + 118 + ], + "ʲ": [ + 119 + ], + "ˈ": [ + 120 + ], + "ˌ": [ + 121 + ], + "ː": [ + 122 + ], + "ˑ": [ + 123 + ], + "˞": [ + 124 + ], + "β": [ + 125 + ], + "θ": [ + 126 + ], + "χ": [ + 127 + ], + "ᵻ": [ + 128 + ], + "ⱱ": [ + 129 + ] + }, + "num_symbols": 130, + "num_speakers": 1, + "speaker_id_map": {}, + "piper_version": "0.2.0", + "language": { + "code": "is_IS", + "family": "is", + "region": "IS", + "name_native": "íslenska", + "name_english": "Icelandic", + "country_english": "Iceland" + }, + "dataset": "salka" +} \ No newline at end of file diff --git a/is/is_IS/salka/medium/samples/speaker_0.mp3 b/is/is_IS/salka/medium/samples/speaker_0.mp3 new file mode 100644 index 0000000000000000000000000000000000000000..df9f4d99a3f61fc4342adcb362076548eed539cc Binary files /dev/null and b/is/is_IS/salka/medium/samples/speaker_0.mp3 differ diff --git a/is/is_IS/steinn/medium/ALIASES b/is/is_IS/steinn/medium/ALIASES new file mode 100644 index 0000000000000000000000000000000000000000..696bcf3983836bc5fc821e1a0de79eaa7156c7fc --- /dev/null +++ b/is/is_IS/steinn/medium/ALIASES @@ -0,0 +1 @@ +is-steinn-medium diff --git a/is/is_IS/steinn/medium/MODEL_CARD b/is/is_IS/steinn/medium/MODEL_CARD new file mode 100644 index 0000000000000000000000000000000000000000..97ee40854e4a578d493ab7f1260782851716c487 --- /dev/null +++ b/is/is_IS/steinn/medium/MODEL_CARD @@ -0,0 +1,15 @@ +# Model card for steinn (medium) + +* Language: is_IS (Icelandic, Iceland) +* Speakers: 1 +* Quality: medium +* Samplerate: 22,050Hz + +## Dataset + +* URL: http://hdl.handle.net/20.500.12537/104 +* License: See URL + +## Training + +Trained from scratch. diff --git a/is/is_IS/steinn/medium/is_IS-steinn-medium.onnx b/is/is_IS/steinn/medium/is_IS-steinn-medium.onnx new file mode 100644 index 0000000000000000000000000000000000000000..796772cc35c482baab0ef636077e58101212751e --- /dev/null +++ b/is/is_IS/steinn/medium/is_IS-steinn-medium.onnx @@ -0,0 +1,3 @@ +version https://git-lfs.github.com/spec/v1 +oid sha256:f10fa173c46fb9b9b683066e3198f61716ab8ab310ffce67004ba25d86bb74b1 +size 76495465 diff --git a/is/is_IS/steinn/medium/is_IS-steinn-medium.onnx.json b/is/is_IS/steinn/medium/is_IS-steinn-medium.onnx.json new file mode 100644 index 0000000000000000000000000000000000000000..4f906935e60f44b1d476a1b51dc4e347da0ccf92 --- /dev/null +++ b/is/is_IS/steinn/medium/is_IS-steinn-medium.onnx.json @@ -0,0 +1,420 @@ +{ + "audio": { + "sample_rate": 22050, + "quality": "medium" + }, + "espeak": { + "voice": "is" + }, + "inference": { + "noise_scale": 0.667, + "length_scale": 1, + "noise_w": 0.8 + }, + "phoneme_map": {}, + "phoneme_id_map": { + "_": [ + 0 + ], + "^": [ + 1 + ], + "$": [ + 2 + ], + " ": [ + 3 + ], + "!": [ + 4 + ], + "'": [ + 5 + ], + "(": [ + 6 + ], + ")": [ + 7 + ], + ",": [ + 8 + ], + "-": [ + 9 + ], + ".": [ + 10 + ], + ":": [ + 11 + ], + ";": [ + 12 + ], + "?": [ + 13 + ], + "a": [ + 14 + ], + "b": [ + 15 + ], + "c": [ + 16 + ], + "d": [ + 17 + ], + "e": [ + 18 + ], + "f": [ + 19 + ], + "h": [ + 20 + ], + "i": [ + 21 + ], + "j": [ + 22 + ], + "k": [ + 23 + ], + "l": [ + 24 + ], + "m": [ + 25 + ], + "n": [ + 26 + ], + "o": [ + 27 + ], + "p": [ + 28 + ], + "q": [ + 29 + ], + "r": [ + 30 + ], + "s": [ + 31 + ], + "t": [ + 32 + ], + "u": [ + 33 + ], + "v": [ + 34 + ], + "w": [ + 35 + ], + "x": [ + 36 + ], + "y": [ + 37 + ], + "z": [ + 38 + ], + "æ": [ + 39 + ], + "ç": [ + 40 + ], + "ð": [ + 41 + ], + "ø": [ + 42 + ], + "ħ": [ + 43 + ], + "ŋ": [ + 44 + ], + "œ": [ + 45 + ], + "ǀ": [ + 46 + ], + "ǁ": [ + 47 + ], + "ǂ": [ + 48 + ], + "ǃ": [ + 49 + ], + "ɐ": [ + 50 + ], + "ɑ": [ + 51 + ], + "ɒ": [ + 52 + ], + "ɓ": [ + 53 + ], + "ɔ": [ + 54 + ], + "ɕ": [ + 55 + ], + "ɖ": [ + 56 + ], + "ɗ": [ + 57 + ], + "ɘ": [ + 58 + ], + "ə": [ + 59 + ], + "ɚ": [ + 60 + ], + "ɛ": [ + 61 + ], + "ɜ": [ + 62 + ], + "ɞ": [ + 63 + ], + "ɟ": [ + 64 + ], + "ɠ": [ + 65 + ], + "ɡ": [ + 66 + ], + "ɢ": [ + 67 + ], + "ɣ": [ + 68 + ], + "ɤ": [ + 69 + ], + "ɥ": [ + 70 + ], + "ɦ": [ + 71 + ], + "ɧ": [ + 72 + ], + "ɨ": [ + 73 + ], + "ɪ": [ + 74 + ], + "ɫ": [ + 75 + ], + "ɬ": [ + 76 + ], + "ɭ": [ + 77 + ], + "ɮ": [ + 78 + ], + "ɯ": [ + 79 + ], + "ɰ": [ + 80 + ], + "ɱ": [ + 81 + ], + "ɲ": [ + 82 + ], + "ɳ": [ + 83 + ], + "ɴ": [ + 84 + ], + "ɵ": [ + 85 + ], + "ɶ": [ + 86 + ], + "ɸ": [ + 87 + ], + "ɹ": [ + 88 + ], + "ɺ": [ + 89 + ], + "ɻ": [ + 90 + ], + "ɽ": [ + 91 + ], + "ɾ": [ + 92 + ], + "ʀ": [ + 93 + ], + "ʁ": [ + 94 + ], + "ʂ": [ + 95 + ], + "ʃ": [ + 96 + ], + "ʄ": [ + 97 + ], + "ʈ": [ + 98 + ], + "ʉ": [ + 99 + ], + "ʊ": [ + 100 + ], + "ʋ": [ + 101 + ], + "ʌ": [ + 102 + ], + "ʍ": [ + 103 + ], + "ʎ": [ + 104 + ], + "ʏ": [ + 105 + ], + "ʐ": [ + 106 + ], + "ʑ": [ + 107 + ], + "ʒ": [ + 108 + ], + "ʔ": [ + 109 + ], + "ʕ": [ + 110 + ], + "ʘ": [ + 111 + ], + "ʙ": [ + 112 + ], + "ʛ": [ + 113 + ], + "ʜ": [ + 114 + ], + "ʝ": [ + 115 + ], + "ʟ": [ + 116 + ], + "ʡ": [ + 117 + ], + "ʢ": [ + 118 + ], + "ʲ": [ + 119 + ], + "ˈ": [ + 120 + ], + "ˌ": [ + 121 + ], + "ː": [ + 122 + ], + "ˑ": [ + 123 + ], + "˞": [ + 124 + ], + "β": [ + 125 + ], + "θ": [ + 126 + ], + "χ": [ + 127 + ], + "ᵻ": [ + 128 + ], + "ⱱ": [ + 129 + ] + }, + "num_symbols": 130, + "num_speakers": 1, + "speaker_id_map": {}, + "piper_version": "0.2.0", + "language": { + "code": "is_IS", + "family": "is", + "region": "IS", + "name_native": "íslenska", + "name_english": "Icelandic", + "country_english": "Iceland" + }, + "dataset": "steinn" +} \ No newline at end of file diff --git a/is/is_IS/steinn/medium/samples/speaker_0.mp3 b/is/is_IS/steinn/medium/samples/speaker_0.mp3 new file mode 100644 index 0000000000000000000000000000000000000000..caf805391cf4d1878005104b9226266d99781bb8 Binary files /dev/null and b/is/is_IS/steinn/medium/samples/speaker_0.mp3 differ diff --git a/is/is_IS/ugla/medium/ALIASES b/is/is_IS/ugla/medium/ALIASES new file mode 100644 index 0000000000000000000000000000000000000000..1353e8c9dc3163d5c7b951bb2ffa0e4480470f23 --- /dev/null +++ b/is/is_IS/ugla/medium/ALIASES @@ -0,0 +1 @@ +is-ugla-medium diff --git a/is/is_IS/ugla/medium/MODEL_CARD b/is/is_IS/ugla/medium/MODEL_CARD new file mode 100644 index 0000000000000000000000000000000000000000..62472b68c4246b7b11492347a8cca25a12e2d986 --- /dev/null +++ b/is/is_IS/ugla/medium/MODEL_CARD @@ -0,0 +1,15 @@ +# Model card for ugla (medium) + +* Language: is_IS (Icelandic, Iceland) +* Speakers: 1 +* Quality: medium +* Samplerate: 22,050Hz + +## Dataset + +* URL: http://hdl.handle.net/20.500.12537/104 +* License: See URL + +## Training + +Trained from scratch. diff --git a/is/is_IS/ugla/medium/is_IS-ugla-medium.onnx b/is/is_IS/ugla/medium/is_IS-ugla-medium.onnx new file mode 100644 index 0000000000000000000000000000000000000000..05e667bd6236026b894ad2c1adbf42b6267ba6a3 --- /dev/null +++ b/is/is_IS/ugla/medium/is_IS-ugla-medium.onnx @@ -0,0 +1,3 @@ +version https://git-lfs.github.com/spec/v1 +oid sha256:b43aef7648d13f68b6db0032d85716b82ca78716218a68442fd7f12444a73b24 +size 76495465 diff --git a/is/is_IS/ugla/medium/is_IS-ugla-medium.onnx.json b/is/is_IS/ugla/medium/is_IS-ugla-medium.onnx.json new file mode 100644 index 0000000000000000000000000000000000000000..2947956a437321f041cbfc75e69a3dcec56a0be1 --- /dev/null +++ b/is/is_IS/ugla/medium/is_IS-ugla-medium.onnx.json @@ -0,0 +1,420 @@ +{ + "audio": { + "sample_rate": 22050, + "quality": "medium" + }, + "espeak": { + "voice": "is" + }, + "inference": { + "noise_scale": 0.667, + "length_scale": 1, + "noise_w": 0.8 + }, + "phoneme_map": {}, + "phoneme_id_map": { + "_": [ + 0 + ], + "^": [ + 1 + ], + "$": [ + 2 + ], + " ": [ + 3 + ], + "!": [ + 4 + ], + "'": [ + 5 + ], + "(": [ + 6 + ], + ")": [ + 7 + ], + ",": [ + 8 + ], + "-": [ + 9 + ], + ".": [ + 10 + ], + ":": [ + 11 + ], + ";": [ + 12 + ], + "?": [ + 13 + ], + "a": [ + 14 + ], + "b": [ + 15 + ], + "c": [ + 16 + ], + "d": [ + 17 + ], + "e": [ + 18 + ], + "f": [ + 19 + ], + "h": [ + 20 + ], + "i": [ + 21 + ], + "j": [ + 22 + ], + "k": [ + 23 + ], + "l": [ + 24 + ], + "m": [ + 25 + ], + "n": [ + 26 + ], + "o": [ + 27 + ], + "p": [ + 28 + ], + "q": [ + 29 + ], + "r": [ + 30 + ], + "s": [ + 31 + ], + "t": [ + 32 + ], + "u": [ + 33 + ], + "v": [ + 34 + ], + "w": [ + 35 + ], + "x": [ + 36 + ], + "y": [ + 37 + ], + "z": [ + 38 + ], + "æ": [ + 39 + ], + "ç": [ + 40 + ], + "ð": [ + 41 + ], + "ø": [ + 42 + ], + "ħ": [ + 43 + ], + "ŋ": [ + 44 + ], + "œ": [ + 45 + ], + "ǀ": [ + 46 + ], + "ǁ": [ + 47 + ], + "ǂ": [ + 48 + ], + "ǃ": [ + 49 + ], + "ɐ": [ + 50 + ], + "ɑ": [ + 51 + ], + "ɒ": [ + 52 + ], + "ɓ": [ + 53 + ], + "ɔ": [ + 54 + ], + "ɕ": [ + 55 + ], + "ɖ": [ + 56 + ], + "ɗ": [ + 57 + ], + "ɘ": [ + 58 + ], + "ə": [ + 59 + ], + "ɚ": [ + 60 + ], + "ɛ": [ + 61 + ], + "ɜ": [ + 62 + ], + "ɞ": [ + 63 + ], + "ɟ": [ + 64 + ], + "ɠ": [ + 65 + ], + "ɡ": [ + 66 + ], + "ɢ": [ + 67 + ], + "ɣ": [ + 68 + ], + "ɤ": [ + 69 + ], + "ɥ": [ + 70 + ], + "ɦ": [ + 71 + ], + "ɧ": [ + 72 + ], + "ɨ": [ + 73 + ], + "ɪ": [ + 74 + ], + "ɫ": [ + 75 + ], + "ɬ": [ + 76 + ], + "ɭ": [ + 77 + ], + "ɮ": [ + 78 + ], + "ɯ": [ + 79 + ], + "ɰ": [ + 80 + ], + "ɱ": [ + 81 + ], + "ɲ": [ + 82 + ], + "ɳ": [ + 83 + ], + "ɴ": [ + 84 + ], + "ɵ": [ + 85 + ], + "ɶ": [ + 86 + ], + "ɸ": [ + 87 + ], + "ɹ": [ + 88 + ], + "ɺ": [ + 89 + ], + "ɻ": [ + 90 + ], + "ɽ": [ + 91 + ], + "ɾ": [ + 92 + ], + "ʀ": [ + 93 + ], + "ʁ": [ + 94 + ], + "ʂ": [ + 95 + ], + "ʃ": [ + 96 + ], + "ʄ": [ + 97 + ], + "ʈ": [ + 98 + ], + "ʉ": [ + 99 + ], + "ʊ": [ + 100 + ], + "ʋ": [ + 101 + ], + "ʌ": [ + 102 + ], + "ʍ": [ + 103 + ], + "ʎ": [ + 104 + ], + "ʏ": [ + 105 + ], + "ʐ": [ + 106 + ], + "ʑ": [ + 107 + ], + "ʒ": [ + 108 + ], + "ʔ": [ + 109 + ], + "ʕ": [ + 110 + ], + "ʘ": [ + 111 + ], + "ʙ": [ + 112 + ], + "ʛ": [ + 113 + ], + "ʜ": [ + 114 + ], + "ʝ": [ + 115 + ], + "ʟ": [ + 116 + ], + "ʡ": [ + 117 + ], + "ʢ": [ + 118 + ], + "ʲ": [ + 119 + ], + "ˈ": [ + 120 + ], + "ˌ": [ + 121 + ], + "ː": [ + 122 + ], + "ˑ": [ + 123 + ], + "˞": [ + 124 + ], + "β": [ + 125 + ], + "θ": [ + 126 + ], + "χ": [ + 127 + ], + "ᵻ": [ + 128 + ], + "ⱱ": [ + 129 + ] + }, + "num_symbols": 130, + "num_speakers": 1, + "speaker_id_map": {}, + "piper_version": "0.2.0", + "language": { + "code": "is_IS", + "family": "is", + "region": "IS", + "name_native": "íslenska", + "name_english": "Icelandic", + "country_english": "Iceland" + }, + "dataset": "ugla" +} \ No newline at end of file diff --git a/is/is_IS/ugla/medium/samples/speaker_0.mp3 b/is/is_IS/ugla/medium/samples/speaker_0.mp3 new file mode 100644 index 0000000000000000000000000000000000000000..922e7fb0f1aef2f3e67c1d7e5c396755e7cf81ba Binary files /dev/null and b/is/is_IS/ugla/medium/samples/speaker_0.mp3 differ diff --git a/it/it_IT/riccardo/x_low/ALIASES b/it/it_IT/riccardo/x_low/ALIASES new file mode 100644 index 0000000000000000000000000000000000000000..ac507729cbadd4bb605c65c12d6acf8a61a210a4 --- /dev/null +++ b/it/it_IT/riccardo/x_low/ALIASES @@ -0,0 +1 @@ +it-riccardo_fasol-x-low diff --git a/it/it_IT/riccardo/x_low/MODEL_CARD b/it/it_IT/riccardo/x_low/MODEL_CARD new file mode 100644 index 0000000000000000000000000000000000000000..4b91ee4189954d51b466372d33c862ef41b70993 --- /dev/null +++ b/it/it_IT/riccardo/x_low/MODEL_CARD @@ -0,0 +1,15 @@ +# Model card for riccardo (x_low) + +* Language: it_IT (Italian, Italy) +* Speakers: 1 +* Quality: x_low +* Samplerate: 16,000Hz + +## Dataset + +* URL: https://www.caito.de/2019/01/03/the-m-ailabs-speech-dataset/ +* License: See URL + +## Training + +Trained from scratch. diff --git a/it/it_IT/riccardo/x_low/it_IT-riccardo-x_low.onnx b/it/it_IT/riccardo/x_low/it_IT-riccardo-x_low.onnx new file mode 100644 index 0000000000000000000000000000000000000000..267840572cc88c433f4482625d8627288394cf28 --- /dev/null +++ b/it/it_IT/riccardo/x_low/it_IT-riccardo-x_low.onnx @@ -0,0 +1,3 @@ +version https://git-lfs.github.com/spec/v1 +oid sha256:1368de15f123275a7ef951c9e5e30be0f58a032daa14a0da44037443c1d1d21b +size 28130791 diff --git a/it/it_IT/riccardo/x_low/it_IT-riccardo-x_low.onnx.json b/it/it_IT/riccardo/x_low/it_IT-riccardo-x_low.onnx.json new file mode 100644 index 0000000000000000000000000000000000000000..8412958723b1007259b0dadee726cf2c00f4606e --- /dev/null +++ b/it/it_IT/riccardo/x_low/it_IT-riccardo-x_low.onnx.json @@ -0,0 +1,420 @@ +{ + "audio": { + "sample_rate": 16000, + "quality": "x_low" + }, + "espeak": { + "voice": "it" + }, + "inference": { + "noise_scale": 0.667, + "length_scale": 1, + "noise_w": 0.8 + }, + "phoneme_map": {}, + "phoneme_id_map": { + "_": [ + 0 + ], + "^": [ + 1 + ], + "$": [ + 2 + ], + " ": [ + 3 + ], + "!": [ + 4 + ], + "'": [ + 5 + ], + "(": [ + 6 + ], + ")": [ + 7 + ], + ",": [ + 8 + ], + "-": [ + 9 + ], + ".": [ + 10 + ], + ":": [ + 11 + ], + ";": [ + 12 + ], + "?": [ + 13 + ], + "a": [ + 14 + ], + "b": [ + 15 + ], + "c": [ + 16 + ], + "d": [ + 17 + ], + "e": [ + 18 + ], + "f": [ + 19 + ], + "h": [ + 20 + ], + "i": [ + 21 + ], + "j": [ + 22 + ], + "k": [ + 23 + ], + "l": [ + 24 + ], + "m": [ + 25 + ], + "n": [ + 26 + ], + "o": [ + 27 + ], + "p": [ + 28 + ], + "q": [ + 29 + ], + "r": [ + 30 + ], + "s": [ + 31 + ], + "t": [ + 32 + ], + "u": [ + 33 + ], + "v": [ + 34 + ], + "w": [ + 35 + ], + "x": [ + 36 + ], + "y": [ + 37 + ], + "z": [ + 38 + ], + "æ": [ + 39 + ], + "ç": [ + 40 + ], + "ð": [ + 41 + ], + "ø": [ + 42 + ], + "ħ": [ + 43 + ], + "ŋ": [ + 44 + ], + "œ": [ + 45 + ], + "ǀ": [ + 46 + ], + "ǁ": [ + 47 + ], + "ǂ": [ + 48 + ], + "ǃ": [ + 49 + ], + "ɐ": [ + 50 + ], + "ɑ": [ + 51 + ], + "ɒ": [ + 52 + ], + "ɓ": [ + 53 + ], + "ɔ": [ + 54 + ], + "ɕ": [ + 55 + ], + "ɖ": [ + 56 + ], + "ɗ": [ + 57 + ], + "ɘ": [ + 58 + ], + "ə": [ + 59 + ], + "ɚ": [ + 60 + ], + "ɛ": [ + 61 + ], + "ɜ": [ + 62 + ], + "ɞ": [ + 63 + ], + "ɟ": [ + 64 + ], + "ɠ": [ + 65 + ], + "ɡ": [ + 66 + ], + "ɢ": [ + 67 + ], + "ɣ": [ + 68 + ], + "ɤ": [ + 69 + ], + "ɥ": [ + 70 + ], + "ɦ": [ + 71 + ], + "ɧ": [ + 72 + ], + "ɨ": [ + 73 + ], + "ɪ": [ + 74 + ], + "ɫ": [ + 75 + ], + "ɬ": [ + 76 + ], + "ɭ": [ + 77 + ], + "ɮ": [ + 78 + ], + "ɯ": [ + 79 + ], + "ɰ": [ + 80 + ], + "ɱ": [ + 81 + ], + "ɲ": [ + 82 + ], + "ɳ": [ + 83 + ], + "ɴ": [ + 84 + ], + "ɵ": [ + 85 + ], + "ɶ": [ + 86 + ], + "ɸ": [ + 87 + ], + "ɹ": [ + 88 + ], + "ɺ": [ + 89 + ], + "ɻ": [ + 90 + ], + "ɽ": [ + 91 + ], + "ɾ": [ + 92 + ], + "ʀ": [ + 93 + ], + "ʁ": [ + 94 + ], + "ʂ": [ + 95 + ], + "ʃ": [ + 96 + ], + "ʄ": [ + 97 + ], + "ʈ": [ + 98 + ], + "ʉ": [ + 99 + ], + "ʊ": [ + 100 + ], + "ʋ": [ + 101 + ], + "ʌ": [ + 102 + ], + "ʍ": [ + 103 + ], + "ʎ": [ + 104 + ], + "ʏ": [ + 105 + ], + "ʐ": [ + 106 + ], + "ʑ": [ + 107 + ], + "ʒ": [ + 108 + ], + "ʔ": [ + 109 + ], + "ʕ": [ + 110 + ], + "ʘ": [ + 111 + ], + "ʙ": [ + 112 + ], + "ʛ": [ + 113 + ], + "ʜ": [ + 114 + ], + "ʝ": [ + 115 + ], + "ʟ": [ + 116 + ], + "ʡ": [ + 117 + ], + "ʢ": [ + 118 + ], + "ʲ": [ + 119 + ], + "ˈ": [ + 120 + ], + "ˌ": [ + 121 + ], + "ː": [ + 122 + ], + "ˑ": [ + 123 + ], + "˞": [ + 124 + ], + "β": [ + 125 + ], + "θ": [ + 126 + ], + "χ": [ + 127 + ], + "ᵻ": [ + 128 + ], + "ⱱ": [ + 129 + ] + }, + "num_symbols": 130, + "num_speakers": 1, + "speaker_id_map": {}, + "piper_version": "0.2.0", + "language": { + "code": "it_IT", + "family": "it", + "region": "IT", + "name_native": "Italiano", + "name_english": "Italian", + "country_english": "Italy" + }, + "dataset": "riccardo" +} \ No newline at end of file diff --git a/it/it_IT/riccardo/x_low/samples/speaker_0.mp3 b/it/it_IT/riccardo/x_low/samples/speaker_0.mp3 new file mode 100644 index 0000000000000000000000000000000000000000..7db17e6110e7010782099acef200aeadac514549 Binary files /dev/null and b/it/it_IT/riccardo/x_low/samples/speaker_0.mp3 differ diff --git a/ka/ka_GE/natia/medium/MODEL_CARD b/ka/ka_GE/natia/medium/MODEL_CARD new file mode 100644 index 0000000000000000000000000000000000000000..37f2267bd1a925188c7291d8d7b3cde03a29e682 --- /dev/null +++ b/ka/ka_GE/natia/medium/MODEL_CARD @@ -0,0 +1,15 @@ +# Model card for natia (medium) + +* Language: ka_GE (Georgian, Georgia) +* Speakers: 1 +* Quality: medium +* Samplerate: 22,050Hz + +## Dataset + +* URL: http://github.com/Olga-Yakovleva/RHVoice +* License: See LICENSE file + +## Training + +Finetuned from U.S. English lessac voice (medium quality). diff --git a/ka/ka_GE/natia/medium/ka_GE-natia-medium.onnx b/ka/ka_GE/natia/medium/ka_GE-natia-medium.onnx new file mode 100644 index 0000000000000000000000000000000000000000..354f9f96da38388907873b8e5f91cdc92b7c4052 --- /dev/null +++ b/ka/ka_GE/natia/medium/ka_GE-natia-medium.onnx @@ -0,0 +1,3 @@ +version https://git-lfs.github.com/spec/v1 +oid sha256:04bdacf188fa24499885f9109b395fe8561a05ec2cd90d55453ec5beed7af460 +size 63201294 diff --git a/ka/ka_GE/natia/medium/ka_GE-natia-medium.onnx.json b/ka/ka_GE/natia/medium/ka_GE-natia-medium.onnx.json new file mode 100644 index 0000000000000000000000000000000000000000..32d055d1cd4652164db21f60980e1cad0add979e --- /dev/null +++ b/ka/ka_GE/natia/medium/ka_GE-natia-medium.onnx.json @@ -0,0 +1,487 @@ +{ + "audio": { + "sample_rate": 22050, + "quality": "medium" + }, + "espeak": { + "voice": "ka" + }, + "inference": { + "noise_scale": 0.667, + "length_scale": 1, + "noise_w": 0.8 + }, + "phoneme_type": "espeak", + "phoneme_map": {}, + "phoneme_id_map": { + "_": [ + 0 + ], + "^": [ + 1 + ], + "$": [ + 2 + ], + " ": [ + 3 + ], + "!": [ + 4 + ], + "'": [ + 5 + ], + "(": [ + 6 + ], + ")": [ + 7 + ], + ",": [ + 8 + ], + "-": [ + 9 + ], + ".": [ + 10 + ], + ":": [ + 11 + ], + ";": [ + 12 + ], + "?": [ + 13 + ], + "a": [ + 14 + ], + "b": [ + 15 + ], + "c": [ + 16 + ], + "d": [ + 17 + ], + "e": [ + 18 + ], + "f": [ + 19 + ], + "h": [ + 20 + ], + "i": [ + 21 + ], + "j": [ + 22 + ], + "k": [ + 23 + ], + "l": [ + 24 + ], + "m": [ + 25 + ], + "n": [ + 26 + ], + "o": [ + 27 + ], + "p": [ + 28 + ], + "q": [ + 29 + ], + "r": [ + 30 + ], + "s": [ + 31 + ], + "t": [ + 32 + ], + "u": [ + 33 + ], + "v": [ + 34 + ], + "w": [ + 35 + ], + "x": [ + 36 + ], + "y": [ + 37 + ], + "z": [ + 38 + ], + "æ": [ + 39 + ], + "ç": [ + 40 + ], + "ð": [ + 41 + ], + "ø": [ + 42 + ], + "ħ": [ + 43 + ], + "ŋ": [ + 44 + ], + "œ": [ + 45 + ], + "ǀ": [ + 46 + ], + "ǁ": [ + 47 + ], + "ǂ": [ + 48 + ], + "ǃ": [ + 49 + ], + "ɐ": [ + 50 + ], + "ɑ": [ + 51 + ], + "ɒ": [ + 52 + ], + "ɓ": [ + 53 + ], + "ɔ": [ + 54 + ], + "ɕ": [ + 55 + ], + "ɖ": [ + 56 + ], + "ɗ": [ + 57 + ], + "ɘ": [ + 58 + ], + "ə": [ + 59 + ], + "ɚ": [ + 60 + ], + "ɛ": [ + 61 + ], + "ɜ": [ + 62 + ], + "ɞ": [ + 63 + ], + "ɟ": [ + 64 + ], + "ɠ": [ + 65 + ], + "ɡ": [ + 66 + ], + "ɢ": [ + 67 + ], + "ɣ": [ + 68 + ], + "ɤ": [ + 69 + ], + "ɥ": [ + 70 + ], + "ɦ": [ + 71 + ], + "ɧ": [ + 72 + ], + "ɨ": [ + 73 + ], + "ɪ": [ + 74 + ], + "ɫ": [ + 75 + ], + "ɬ": [ + 76 + ], + "ɭ": [ + 77 + ], + "ɮ": [ + 78 + ], + "ɯ": [ + 79 + ], + "ɰ": [ + 80 + ], + "ɱ": [ + 81 + ], + "ɲ": [ + 82 + ], + "ɳ": [ + 83 + ], + "ɴ": [ + 84 + ], + "ɵ": [ + 85 + ], + "ɶ": [ + 86 + ], + "ɸ": [ + 87 + ], + "ɹ": [ + 88 + ], + "ɺ": [ + 89 + ], + "ɻ": [ + 90 + ], + "ɽ": [ + 91 + ], + "ɾ": [ + 92 + ], + "ʀ": [ + 93 + ], + "ʁ": [ + 94 + ], + "ʂ": [ + 95 + ], + "ʃ": [ + 96 + ], + "ʄ": [ + 97 + ], + "ʈ": [ + 98 + ], + "ʉ": [ + 99 + ], + "ʊ": [ + 100 + ], + "ʋ": [ + 101 + ], + "ʌ": [ + 102 + ], + "ʍ": [ + 103 + ], + "ʎ": [ + 104 + ], + "ʏ": [ + 105 + ], + "ʐ": [ + 106 + ], + "ʑ": [ + 107 + ], + "ʒ": [ + 108 + ], + "ʔ": [ + 109 + ], + "ʕ": [ + 110 + ], + "ʘ": [ + 111 + ], + "ʙ": [ + 112 + ], + "ʛ": [ + 113 + ], + "ʜ": [ + 114 + ], + "ʝ": [ + 115 + ], + "ʟ": [ + 116 + ], + "ʡ": [ + 117 + ], + "ʢ": [ + 118 + ], + "ʲ": [ + 119 + ], + "ˈ": [ + 120 + ], + "ˌ": [ + 121 + ], + "ː": [ + 122 + ], + "ˑ": [ + 123 + ], + "˞": [ + 124 + ], + "β": [ + 125 + ], + "θ": [ + 126 + ], + "χ": [ + 127 + ], + "ᵻ": [ + 128 + ], + "ⱱ": [ + 129 + ], + "0": [ + 130 + ], + "1": [ + 131 + ], + "2": [ + 132 + ], + "3": [ + 133 + ], + "4": [ + 134 + ], + "5": [ + 135 + ], + "6": [ + 136 + ], + "7": [ + 137 + ], + "8": [ + 138 + ], + "9": [ + 139 + ], + "̧": [ + 140 + ], + "̃": [ + 141 + ], + "̪": [ + 142 + ], + "̯": [ + 143 + ], + "̩": [ + 144 + ], + "ʰ": [ + 145 + ], + "ˤ": [ + 146 + ], + "ε": [ + 147 + ], + "↓": [ + 148 + ], + "#": [ + 149 + ], + "\"": [ + 150 + ], + "↑": [ + 151 + ] + }, + "num_symbols": 256, + "num_speakers": 1, + "speaker_id_map": {}, + "piper_version": "1.0.0", + "language": { + "code": "ka_GE", + "family": "ka", + "region": "GE", + "name_native": "ქართული ენა", + "name_english": "Georgian", + "country_english": "Georgia" + }, + "dataset": "natia" +} \ No newline at end of file diff --git a/ka/ka_GE/natia/medium/samples/speaker_0.mp3 b/ka/ka_GE/natia/medium/samples/speaker_0.mp3 new file mode 100644 index 0000000000000000000000000000000000000000..392e9fdebc5ab6cb53723210dd1e87d8aa6c1bfb Binary files /dev/null and b/ka/ka_GE/natia/medium/samples/speaker_0.mp3 differ diff --git a/kk/kk_KZ/iseke/x_low/ALIASES b/kk/kk_KZ/iseke/x_low/ALIASES new file mode 100644 index 0000000000000000000000000000000000000000..412a89acb68bec894b7913c7d8ff7ad5a82b39ea --- /dev/null +++ b/kk/kk_KZ/iseke/x_low/ALIASES @@ -0,0 +1 @@ +kk-iseke-x-low diff --git a/kk/kk_KZ/iseke/x_low/MODEL_CARD b/kk/kk_KZ/iseke/x_low/MODEL_CARD new file mode 100644 index 0000000000000000000000000000000000000000..f53968ae19def9449efb653a7c49783716b256f1 --- /dev/null +++ b/kk/kk_KZ/iseke/x_low/MODEL_CARD @@ -0,0 +1,15 @@ +# Model card for iseke (x_low) + +* Language: kk_KZ (Kazakh, Kazakhstan) +* Speakers: 1 +* Quality: x_low +* Samplerate: 16,000Hz + +## Dataset + +* URL: https://github.com/IS2AI/Kazakh_TTS +* License: CC-BY-4.0 + +## Training + +Trained from scratch. diff --git a/kk/kk_KZ/iseke/x_low/kk_KZ-iseke-x_low.onnx b/kk/kk_KZ/iseke/x_low/kk_KZ-iseke-x_low.onnx new file mode 100644 index 0000000000000000000000000000000000000000..8e5264d117f0eec8e329c1ba8eb9c220faae662c --- /dev/null +++ b/kk/kk_KZ/iseke/x_low/kk_KZ-iseke-x_low.onnx @@ -0,0 +1,3 @@ +version https://git-lfs.github.com/spec/v1 +oid sha256:dcf526e1c089823c9f896f8f7fddd9b05d538752e139c8a2b4cf4fba41c9715c +size 28130791 diff --git a/kk/kk_KZ/iseke/x_low/kk_KZ-iseke-x_low.onnx.json b/kk/kk_KZ/iseke/x_low/kk_KZ-iseke-x_low.onnx.json new file mode 100644 index 0000000000000000000000000000000000000000..e27598aa9fec43a2e8af7c6f93557cb58c2db2d4 --- /dev/null +++ b/kk/kk_KZ/iseke/x_low/kk_KZ-iseke-x_low.onnx.json @@ -0,0 +1,420 @@ +{ + "audio": { + "sample_rate": 16000, + "quality": "x_low" + }, + "espeak": { + "voice": "kk" + }, + "inference": { + "noise_scale": 0.667, + "length_scale": 1, + "noise_w": 0.8 + }, + "phoneme_map": {}, + "phoneme_id_map": { + "_": [ + 0 + ], + "^": [ + 1 + ], + "$": [ + 2 + ], + " ": [ + 3 + ], + "!": [ + 4 + ], + "'": [ + 5 + ], + "(": [ + 6 + ], + ")": [ + 7 + ], + ",": [ + 8 + ], + "-": [ + 9 + ], + ".": [ + 10 + ], + ":": [ + 11 + ], + ";": [ + 12 + ], + "?": [ + 13 + ], + "a": [ + 14 + ], + "b": [ + 15 + ], + "c": [ + 16 + ], + "d": [ + 17 + ], + "e": [ + 18 + ], + "f": [ + 19 + ], + "h": [ + 20 + ], + "i": [ + 21 + ], + "j": [ + 22 + ], + "k": [ + 23 + ], + "l": [ + 24 + ], + "m": [ + 25 + ], + "n": [ + 26 + ], + "o": [ + 27 + ], + "p": [ + 28 + ], + "q": [ + 29 + ], + "r": [ + 30 + ], + "s": [ + 31 + ], + "t": [ + 32 + ], + "u": [ + 33 + ], + "v": [ + 34 + ], + "w": [ + 35 + ], + "x": [ + 36 + ], + "y": [ + 37 + ], + "z": [ + 38 + ], + "æ": [ + 39 + ], + "ç": [ + 40 + ], + "ð": [ + 41 + ], + "ø": [ + 42 + ], + "ħ": [ + 43 + ], + "ŋ": [ + 44 + ], + "œ": [ + 45 + ], + "ǀ": [ + 46 + ], + "ǁ": [ + 47 + ], + "ǂ": [ + 48 + ], + "ǃ": [ + 49 + ], + "ɐ": [ + 50 + ], + "ɑ": [ + 51 + ], + "ɒ": [ + 52 + ], + "ɓ": [ + 53 + ], + "ɔ": [ + 54 + ], + "ɕ": [ + 55 + ], + "ɖ": [ + 56 + ], + "ɗ": [ + 57 + ], + "ɘ": [ + 58 + ], + "ə": [ + 59 + ], + "ɚ": [ + 60 + ], + "ɛ": [ + 61 + ], + "ɜ": [ + 62 + ], + "ɞ": [ + 63 + ], + "ɟ": [ + 64 + ], + "ɠ": [ + 65 + ], + "ɡ": [ + 66 + ], + "ɢ": [ + 67 + ], + "ɣ": [ + 68 + ], + "ɤ": [ + 69 + ], + "ɥ": [ + 70 + ], + "ɦ": [ + 71 + ], + "ɧ": [ + 72 + ], + "ɨ": [ + 73 + ], + "ɪ": [ + 74 + ], + "ɫ": [ + 75 + ], + "ɬ": [ + 76 + ], + "ɭ": [ + 77 + ], + "ɮ": [ + 78 + ], + "ɯ": [ + 79 + ], + "ɰ": [ + 80 + ], + "ɱ": [ + 81 + ], + "ɲ": [ + 82 + ], + "ɳ": [ + 83 + ], + "ɴ": [ + 84 + ], + "ɵ": [ + 85 + ], + "ɶ": [ + 86 + ], + "ɸ": [ + 87 + ], + "ɹ": [ + 88 + ], + "ɺ": [ + 89 + ], + "ɻ": [ + 90 + ], + "ɽ": [ + 91 + ], + "ɾ": [ + 92 + ], + "ʀ": [ + 93 + ], + "ʁ": [ + 94 + ], + "ʂ": [ + 95 + ], + "ʃ": [ + 96 + ], + "ʄ": [ + 97 + ], + "ʈ": [ + 98 + ], + "ʉ": [ + 99 + ], + "ʊ": [ + 100 + ], + "ʋ": [ + 101 + ], + "ʌ": [ + 102 + ], + "ʍ": [ + 103 + ], + "ʎ": [ + 104 + ], + "ʏ": [ + 105 + ], + "ʐ": [ + 106 + ], + "ʑ": [ + 107 + ], + "ʒ": [ + 108 + ], + "ʔ": [ + 109 + ], + "ʕ": [ + 110 + ], + "ʘ": [ + 111 + ], + "ʙ": [ + 112 + ], + "ʛ": [ + 113 + ], + "ʜ": [ + 114 + ], + "ʝ": [ + 115 + ], + "ʟ": [ + 116 + ], + "ʡ": [ + 117 + ], + "ʢ": [ + 118 + ], + "ʲ": [ + 119 + ], + "ˈ": [ + 120 + ], + "ˌ": [ + 121 + ], + "ː": [ + 122 + ], + "ˑ": [ + 123 + ], + "˞": [ + 124 + ], + "β": [ + 125 + ], + "θ": [ + 126 + ], + "χ": [ + 127 + ], + "ᵻ": [ + 128 + ], + "ⱱ": [ + 129 + ] + }, + "num_symbols": 130, + "num_speakers": 1, + "speaker_id_map": {}, + "piper_version": "0.2.0", + "language": { + "code": "kk_KZ", + "family": "kk", + "region": "KZ", + "name_native": "қазақша", + "name_english": "Kazakh", + "country_english": "Kazakhstan" + }, + "dataset": "iseke" +} \ No newline at end of file diff --git a/kk/kk_KZ/iseke/x_low/samples/speaker_0.mp3 b/kk/kk_KZ/iseke/x_low/samples/speaker_0.mp3 new file mode 100644 index 0000000000000000000000000000000000000000..96d45d656ee60c7dc07911256a8cc6b0568b8155 Binary files /dev/null and b/kk/kk_KZ/iseke/x_low/samples/speaker_0.mp3 differ diff --git a/kk/kk_KZ/issai/high/ALIASES b/kk/kk_KZ/issai/high/ALIASES new file mode 100644 index 0000000000000000000000000000000000000000..5c7e2d7b02a02d6a398c62a83ed1db1e42f11c64 --- /dev/null +++ b/kk/kk_KZ/issai/high/ALIASES @@ -0,0 +1 @@ +kk-issai-high diff --git a/kk/kk_KZ/issai/high/MODEL_CARD b/kk/kk_KZ/issai/high/MODEL_CARD new file mode 100644 index 0000000000000000000000000000000000000000..7f44fcb74006d58f244aeceecf68cf139fd73c7e --- /dev/null +++ b/kk/kk_KZ/issai/high/MODEL_CARD @@ -0,0 +1,15 @@ +# Model card for issai (high) + +* Language: kk_KZ (Kazakh, Kazakhstan) +* Speakers: 6 +* Quality: high +* Samplerate: 22,050Hz + +## Dataset + +* URL: https://github.com/IS2AI/Kazakh_TTS +* License: CC-BY-4.0 + +## Training + +Trained from scratch. diff --git a/kk/kk_KZ/issai/high/kk_KZ-issai-high.onnx b/kk/kk_KZ/issai/high/kk_KZ-issai-high.onnx new file mode 100644 index 0000000000000000000000000000000000000000..13b88f5b547849b2387ba8cb349f66c73d29c664 --- /dev/null +++ b/kk/kk_KZ/issai/high/kk_KZ-issai-high.onnx @@ -0,0 +1,3 @@ +version https://git-lfs.github.com/spec/v1 +oid sha256:4dee767c893e8535da821447d12cb030e3569e11254c14030a1da5d8b2222c16 +size 127864258 diff --git a/kk/kk_KZ/issai/high/kk_KZ-issai-high.onnx.json b/kk/kk_KZ/issai/high/kk_KZ-issai-high.onnx.json new file mode 100644 index 0000000000000000000000000000000000000000..e49a90994ec9f689f69cdc6418d853c2b4919a95 --- /dev/null +++ b/kk/kk_KZ/issai/high/kk_KZ-issai-high.onnx.json @@ -0,0 +1,427 @@ +{ + "audio": { + "sample_rate": 22050, + "quality": "high" + }, + "espeak": { + "voice": "kk" + }, + "inference": { + "noise_scale": 0.667, + "length_scale": 1, + "noise_w": 0.8 + }, + "phoneme_map": {}, + "phoneme_id_map": { + "_": [ + 0 + ], + "^": [ + 1 + ], + "$": [ + 2 + ], + " ": [ + 3 + ], + "!": [ + 4 + ], + "'": [ + 5 + ], + "(": [ + 6 + ], + ")": [ + 7 + ], + ",": [ + 8 + ], + "-": [ + 9 + ], + ".": [ + 10 + ], + ":": [ + 11 + ], + ";": [ + 12 + ], + "?": [ + 13 + ], + "a": [ + 14 + ], + "b": [ + 15 + ], + "c": [ + 16 + ], + "d": [ + 17 + ], + "e": [ + 18 + ], + "f": [ + 19 + ], + "h": [ + 20 + ], + "i": [ + 21 + ], + "j": [ + 22 + ], + "k": [ + 23 + ], + "l": [ + 24 + ], + "m": [ + 25 + ], + "n": [ + 26 + ], + "o": [ + 27 + ], + "p": [ + 28 + ], + "q": [ + 29 + ], + "r": [ + 30 + ], + "s": [ + 31 + ], + "t": [ + 32 + ], + "u": [ + 33 + ], + "v": [ + 34 + ], + "w": [ + 35 + ], + "x": [ + 36 + ], + "y": [ + 37 + ], + "z": [ + 38 + ], + "æ": [ + 39 + ], + "ç": [ + 40 + ], + "ð": [ + 41 + ], + "ø": [ + 42 + ], + "ħ": [ + 43 + ], + "ŋ": [ + 44 + ], + "œ": [ + 45 + ], + "ǀ": [ + 46 + ], + "ǁ": [ + 47 + ], + "ǂ": [ + 48 + ], + "ǃ": [ + 49 + ], + "ɐ": [ + 50 + ], + "ɑ": [ + 51 + ], + "ɒ": [ + 52 + ], + "ɓ": [ + 53 + ], + "ɔ": [ + 54 + ], + "ɕ": [ + 55 + ], + "ɖ": [ + 56 + ], + "ɗ": [ + 57 + ], + "ɘ": [ + 58 + ], + "ə": [ + 59 + ], + "ɚ": [ + 60 + ], + "ɛ": [ + 61 + ], + "ɜ": [ + 62 + ], + "ɞ": [ + 63 + ], + "ɟ": [ + 64 + ], + "ɠ": [ + 65 + ], + "ɡ": [ + 66 + ], + "ɢ": [ + 67 + ], + "ɣ": [ + 68 + ], + "ɤ": [ + 69 + ], + "ɥ": [ + 70 + ], + "ɦ": [ + 71 + ], + "ɧ": [ + 72 + ], + "ɨ": [ + 73 + ], + "ɪ": [ + 74 + ], + "ɫ": [ + 75 + ], + "ɬ": [ + 76 + ], + "ɭ": [ + 77 + ], + "ɮ": [ + 78 + ], + "ɯ": [ + 79 + ], + "ɰ": [ + 80 + ], + "ɱ": [ + 81 + ], + "ɲ": [ + 82 + ], + "ɳ": [ + 83 + ], + "ɴ": [ + 84 + ], + "ɵ": [ + 85 + ], + "ɶ": [ + 86 + ], + "ɸ": [ + 87 + ], + "ɹ": [ + 88 + ], + "ɺ": [ + 89 + ], + "ɻ": [ + 90 + ], + "ɽ": [ + 91 + ], + "ɾ": [ + 92 + ], + "ʀ": [ + 93 + ], + "ʁ": [ + 94 + ], + "ʂ": [ + 95 + ], + "ʃ": [ + 96 + ], + "ʄ": [ + 97 + ], + "ʈ": [ + 98 + ], + "ʉ": [ + 99 + ], + "ʊ": [ + 100 + ], + "ʋ": [ + 101 + ], + "ʌ": [ + 102 + ], + "ʍ": [ + 103 + ], + "ʎ": [ + 104 + ], + "ʏ": [ + 105 + ], + "ʐ": [ + 106 + ], + "ʑ": [ + 107 + ], + "ʒ": [ + 108 + ], + "ʔ": [ + 109 + ], + "ʕ": [ + 110 + ], + "ʘ": [ + 111 + ], + "ʙ": [ + 112 + ], + "ʛ": [ + 113 + ], + "ʜ": [ + 114 + ], + "ʝ": [ + 115 + ], + "ʟ": [ + 116 + ], + "ʡ": [ + 117 + ], + "ʢ": [ + 118 + ], + "ʲ": [ + 119 + ], + "ˈ": [ + 120 + ], + "ˌ": [ + 121 + ], + "ː": [ + 122 + ], + "ˑ": [ + 123 + ], + "˞": [ + 124 + ], + "β": [ + 125 + ], + "θ": [ + 126 + ], + "χ": [ + 127 + ], + "ᵻ": [ + 128 + ], + "ⱱ": [ + 129 + ] + }, + "num_symbols": 130, + "num_speakers": 6, + "speaker_id_map": { + "ISSAI_KazakhTTS2_M2": 0, + "ISSAI_KazakhTTS_M1_Iseke": 1, + "ISSAI_KazakhTTS2_F3": 2, + "ISSAI_KazakhTTS_F1_Raya": 3, + "ISSAI_KazakhTTS2_F1": 4, + "ISSAI_KazakhTTS2_F2": 5 + }, + "piper_version": "0.2.0", + "language": { + "code": "kk_KZ", + "family": "kk", + "region": "KZ", + "name_native": "қазақша", + "name_english": "Kazakh", + "country_english": "Kazakhstan" + }, + "dataset": "issai" +} \ No newline at end of file diff --git a/kk/kk_KZ/issai/high/samples/speaker_0.mp3 b/kk/kk_KZ/issai/high/samples/speaker_0.mp3 new file mode 100644 index 0000000000000000000000000000000000000000..8580dc0f1023b34f27de8aa889bcb9ad275e0026 Binary files /dev/null and b/kk/kk_KZ/issai/high/samples/speaker_0.mp3 differ diff --git a/kk/kk_KZ/issai/high/samples/speaker_1.mp3 b/kk/kk_KZ/issai/high/samples/speaker_1.mp3 new file mode 100644 index 0000000000000000000000000000000000000000..ddff2cf49bea2f5d19b417e81eb85ee29f5fb316 Binary files /dev/null and b/kk/kk_KZ/issai/high/samples/speaker_1.mp3 differ diff --git a/kk/kk_KZ/issai/high/samples/speaker_2.mp3 b/kk/kk_KZ/issai/high/samples/speaker_2.mp3 new file mode 100644 index 0000000000000000000000000000000000000000..76ccbf4199d0e08cc3cbdc0a588b28785dc10468 Binary files /dev/null and b/kk/kk_KZ/issai/high/samples/speaker_2.mp3 differ diff --git a/kk/kk_KZ/issai/high/samples/speaker_3.mp3 b/kk/kk_KZ/issai/high/samples/speaker_3.mp3 new file mode 100644 index 0000000000000000000000000000000000000000..62f3e0f5f57e4fbd1ee9a8e966420d82166f9449 Binary files /dev/null and b/kk/kk_KZ/issai/high/samples/speaker_3.mp3 differ diff --git a/kk/kk_KZ/issai/high/samples/speaker_4.mp3 b/kk/kk_KZ/issai/high/samples/speaker_4.mp3 new file mode 100644 index 0000000000000000000000000000000000000000..be0c39ba99f53fd23198c84ca97889cc2941127f Binary files /dev/null and b/kk/kk_KZ/issai/high/samples/speaker_4.mp3 differ diff --git a/kk/kk_KZ/issai/high/samples/speaker_5.mp3 b/kk/kk_KZ/issai/high/samples/speaker_5.mp3 new file mode 100644 index 0000000000000000000000000000000000000000..299c9ffbed01a6057065cd509caaecd8769c624d Binary files /dev/null and b/kk/kk_KZ/issai/high/samples/speaker_5.mp3 differ diff --git a/kk/kk_KZ/raya/x_low/ALIASES b/kk/kk_KZ/raya/x_low/ALIASES new file mode 100644 index 0000000000000000000000000000000000000000..17c46f8d4dae1315291de50345c8330f1ec2a9e4 --- /dev/null +++ b/kk/kk_KZ/raya/x_low/ALIASES @@ -0,0 +1 @@ +kk-raya-x-low diff --git a/kk/kk_KZ/raya/x_low/MODEL_CARD b/kk/kk_KZ/raya/x_low/MODEL_CARD new file mode 100644 index 0000000000000000000000000000000000000000..b98b7b4993639908f910d076019c0fce87d7a326 --- /dev/null +++ b/kk/kk_KZ/raya/x_low/MODEL_CARD @@ -0,0 +1,15 @@ +# Model card for raya (x_low) + +* Language: kk_KZ (Kazakh, Kazakhstan) +* Speakers: 1 +* Quality: x_low +* Samplerate: 16,000Hz + +## Dataset + +* URL: https://github.com/IS2AI/Kazakh_TTS +* License: CC-BY-4.0 + +## Training + +Trained from scratch. diff --git a/kk/kk_KZ/raya/x_low/kk_KZ-raya-x_low.onnx b/kk/kk_KZ/raya/x_low/kk_KZ-raya-x_low.onnx new file mode 100644 index 0000000000000000000000000000000000000000..d8594cf3bea112fdc6e23d26b4cc77b1e7e0e513 --- /dev/null +++ b/kk/kk_KZ/raya/x_low/kk_KZ-raya-x_low.onnx @@ -0,0 +1,3 @@ +version https://git-lfs.github.com/spec/v1 +oid sha256:75552af00254baee9f85b976877e039f9530ea86ddb83d70262b72925785a0aa +size 28130791 diff --git a/kk/kk_KZ/raya/x_low/kk_KZ-raya-x_low.onnx.json b/kk/kk_KZ/raya/x_low/kk_KZ-raya-x_low.onnx.json new file mode 100644 index 0000000000000000000000000000000000000000..af499f3989f65b1d37431459c9882e342a85de5f --- /dev/null +++ b/kk/kk_KZ/raya/x_low/kk_KZ-raya-x_low.onnx.json @@ -0,0 +1,420 @@ +{ + "audio": { + "sample_rate": 16000, + "quality": "x_low" + }, + "espeak": { + "voice": "kk" + }, + "inference": { + "noise_scale": 0.667, + "length_scale": 1, + "noise_w": 0.8 + }, + "phoneme_map": {}, + "phoneme_id_map": { + "_": [ + 0 + ], + "^": [ + 1 + ], + "$": [ + 2 + ], + " ": [ + 3 + ], + "!": [ + 4 + ], + "'": [ + 5 + ], + "(": [ + 6 + ], + ")": [ + 7 + ], + ",": [ + 8 + ], + "-": [ + 9 + ], + ".": [ + 10 + ], + ":": [ + 11 + ], + ";": [ + 12 + ], + "?": [ + 13 + ], + "a": [ + 14 + ], + "b": [ + 15 + ], + "c": [ + 16 + ], + "d": [ + 17 + ], + "e": [ + 18 + ], + "f": [ + 19 + ], + "h": [ + 20 + ], + "i": [ + 21 + ], + "j": [ + 22 + ], + "k": [ + 23 + ], + "l": [ + 24 + ], + "m": [ + 25 + ], + "n": [ + 26 + ], + "o": [ + 27 + ], + "p": [ + 28 + ], + "q": [ + 29 + ], + "r": [ + 30 + ], + "s": [ + 31 + ], + "t": [ + 32 + ], + "u": [ + 33 + ], + "v": [ + 34 + ], + "w": [ + 35 + ], + "x": [ + 36 + ], + "y": [ + 37 + ], + "z": [ + 38 + ], + "æ": [ + 39 + ], + "ç": [ + 40 + ], + "ð": [ + 41 + ], + "ø": [ + 42 + ], + "ħ": [ + 43 + ], + "ŋ": [ + 44 + ], + "œ": [ + 45 + ], + "ǀ": [ + 46 + ], + "ǁ": [ + 47 + ], + "ǂ": [ + 48 + ], + "ǃ": [ + 49 + ], + "ɐ": [ + 50 + ], + "ɑ": [ + 51 + ], + "ɒ": [ + 52 + ], + "ɓ": [ + 53 + ], + "ɔ": [ + 54 + ], + "ɕ": [ + 55 + ], + "ɖ": [ + 56 + ], + "ɗ": [ + 57 + ], + "ɘ": [ + 58 + ], + "ə": [ + 59 + ], + "ɚ": [ + 60 + ], + "ɛ": [ + 61 + ], + "ɜ": [ + 62 + ], + "ɞ": [ + 63 + ], + "ɟ": [ + 64 + ], + "ɠ": [ + 65 + ], + "ɡ": [ + 66 + ], + "ɢ": [ + 67 + ], + "ɣ": [ + 68 + ], + "ɤ": [ + 69 + ], + "ɥ": [ + 70 + ], + "ɦ": [ + 71 + ], + "ɧ": [ + 72 + ], + "ɨ": [ + 73 + ], + "ɪ": [ + 74 + ], + "ɫ": [ + 75 + ], + "ɬ": [ + 76 + ], + "ɭ": [ + 77 + ], + "ɮ": [ + 78 + ], + "ɯ": [ + 79 + ], + "ɰ": [ + 80 + ], + "ɱ": [ + 81 + ], + "ɲ": [ + 82 + ], + "ɳ": [ + 83 + ], + "ɴ": [ + 84 + ], + "ɵ": [ + 85 + ], + "ɶ": [ + 86 + ], + "ɸ": [ + 87 + ], + "ɹ": [ + 88 + ], + "ɺ": [ + 89 + ], + "ɻ": [ + 90 + ], + "ɽ": [ + 91 + ], + "ɾ": [ + 92 + ], + "ʀ": [ + 93 + ], + "ʁ": [ + 94 + ], + "ʂ": [ + 95 + ], + "ʃ": [ + 96 + ], + "ʄ": [ + 97 + ], + "ʈ": [ + 98 + ], + "ʉ": [ + 99 + ], + "ʊ": [ + 100 + ], + "ʋ": [ + 101 + ], + "ʌ": [ + 102 + ], + "ʍ": [ + 103 + ], + "ʎ": [ + 104 + ], + "ʏ": [ + 105 + ], + "ʐ": [ + 106 + ], + "ʑ": [ + 107 + ], + "ʒ": [ + 108 + ], + "ʔ": [ + 109 + ], + "ʕ": [ + 110 + ], + "ʘ": [ + 111 + ], + "ʙ": [ + 112 + ], + "ʛ": [ + 113 + ], + "ʜ": [ + 114 + ], + "ʝ": [ + 115 + ], + "ʟ": [ + 116 + ], + "ʡ": [ + 117 + ], + "ʢ": [ + 118 + ], + "ʲ": [ + 119 + ], + "ˈ": [ + 120 + ], + "ˌ": [ + 121 + ], + "ː": [ + 122 + ], + "ˑ": [ + 123 + ], + "˞": [ + 124 + ], + "β": [ + 125 + ], + "θ": [ + 126 + ], + "χ": [ + 127 + ], + "ᵻ": [ + 128 + ], + "ⱱ": [ + 129 + ] + }, + "num_symbols": 130, + "num_speakers": 1, + "speaker_id_map": {}, + "piper_version": "0.2.0", + "language": { + "code": "kk_KZ", + "family": "kk", + "region": "KZ", + "name_native": "қазақша", + "name_english": "Kazakh", + "country_english": "Kazakhstan" + }, + "dataset": "raya" +} \ No newline at end of file diff --git a/kk/kk_KZ/raya/x_low/samples/speaker_0.mp3 b/kk/kk_KZ/raya/x_low/samples/speaker_0.mp3 new file mode 100644 index 0000000000000000000000000000000000000000..b68ec76d56916b6252d8a1189c4a0c6f883b991f Binary files /dev/null and b/kk/kk_KZ/raya/x_low/samples/speaker_0.mp3 differ diff --git a/lb/lb_LU/marylux/medium/MODEL_CARD b/lb/lb_LU/marylux/medium/MODEL_CARD new file mode 100644 index 0000000000000000000000000000000000000000..a9cf58deae873bdba8af4294ffc029bc4c7b69b1 --- /dev/null +++ b/lb/lb_LU/marylux/medium/MODEL_CARD @@ -0,0 +1,15 @@ +# Model card for marylux (medium) + +* Language: lb_LU (Luxembourgish, Luxembourg) +* Speakers: 1 +* Quality: medium +* Samplerate: 22,050Hz + +## Dataset + +* URL: https://github.com/marytts/marylux-data +* License: http://creativecommons.org/licenses/by-nc-sa/4.0/ + +## Training + +Finetuned from U.S. English lessac voice (medium quality). diff --git a/lb/lb_LU/marylux/medium/lb_LU-marylux-medium.onnx b/lb/lb_LU/marylux/medium/lb_LU-marylux-medium.onnx new file mode 100644 index 0000000000000000000000000000000000000000..5eabef72e631b3c579d6896c3d137ba70d2b7110 --- /dev/null +++ b/lb/lb_LU/marylux/medium/lb_LU-marylux-medium.onnx @@ -0,0 +1,3 @@ +version https://git-lfs.github.com/spec/v1 +oid sha256:4147ecacdd98932951d0f956555542de358d3ccff708d4996e305c3ce287097a +size 63201294 diff --git a/lb/lb_LU/marylux/medium/lb_LU-marylux-medium.onnx.json b/lb/lb_LU/marylux/medium/lb_LU-marylux-medium.onnx.json new file mode 100644 index 0000000000000000000000000000000000000000..80fbf06d2cb641becfed4b481444b0357759d3a1 --- /dev/null +++ b/lb/lb_LU/marylux/medium/lb_LU-marylux-medium.onnx.json @@ -0,0 +1,502 @@ +{ + "piper_version": "1.0.0", + "audio": { + "sample_rate": 22050, + "quality": "medium" + }, + "espeak": { + "voice": "lb" + }, + "inference": { + "noise_scale": 0.667, + "length_scale": 1, + "noise_w": 0.8 + }, + "phoneme_type": "espeak", + "phoneme_map": {}, + "phoneme_id_map": { + " ": [ + 3 + ], + "!": [ + 4 + ], + "\"": [ + 150 + ], + "#": [ + 149 + ], + "$": [ + 2 + ], + "'": [ + 5 + ], + "(": [ + 6 + ], + ")": [ + 7 + ], + ",": [ + 8 + ], + "-": [ + 9 + ], + ".": [ + 10 + ], + "0": [ + 130 + ], + "1": [ + 131 + ], + "2": [ + 132 + ], + "3": [ + 133 + ], + "4": [ + 134 + ], + "5": [ + 135 + ], + "6": [ + 136 + ], + "7": [ + 137 + ], + "8": [ + 138 + ], + "9": [ + 139 + ], + ":": [ + 11 + ], + ";": [ + 12 + ], + "?": [ + 13 + ], + "X": [ + 156 + ], + "^": [ + 1 + ], + "_": [ + 0 + ], + "a": [ + 14 + ], + "b": [ + 15 + ], + "c": [ + 16 + ], + "d": [ + 17 + ], + "e": [ + 18 + ], + "f": [ + 19 + ], + "g": [ + 154 + ], + "h": [ + 20 + ], + "i": [ + 21 + ], + "j": [ + 22 + ], + "k": [ + 23 + ], + "l": [ + 24 + ], + "m": [ + 25 + ], + "n": [ + 26 + ], + "o": [ + 27 + ], + "p": [ + 28 + ], + "q": [ + 29 + ], + "r": [ + 30 + ], + "s": [ + 31 + ], + "t": [ + 32 + ], + "u": [ + 33 + ], + "v": [ + 34 + ], + "w": [ + 35 + ], + "x": [ + 36 + ], + "y": [ + 37 + ], + "z": [ + 38 + ], + "æ": [ + 39 + ], + "ç": [ + 40 + ], + "ð": [ + 41 + ], + "ø": [ + 42 + ], + "ħ": [ + 43 + ], + "ŋ": [ + 44 + ], + "œ": [ + 45 + ], + "ǀ": [ + 46 + ], + "ǁ": [ + 47 + ], + "ǂ": [ + 48 + ], + "ǃ": [ + 49 + ], + "ɐ": [ + 50 + ], + "ɑ": [ + 51 + ], + "ɒ": [ + 52 + ], + "ɓ": [ + 53 + ], + "ɔ": [ + 54 + ], + "ɕ": [ + 55 + ], + "ɖ": [ + 56 + ], + "ɗ": [ + 57 + ], + "ɘ": [ + 58 + ], + "ə": [ + 59 + ], + "ɚ": [ + 60 + ], + "ɛ": [ + 61 + ], + "ɜ": [ + 62 + ], + "ɞ": [ + 63 + ], + "ɟ": [ + 64 + ], + "ɠ": [ + 65 + ], + "ɡ": [ + 66 + ], + "ɢ": [ + 67 + ], + "ɣ": [ + 68 + ], + "ɤ": [ + 69 + ], + "ɥ": [ + 70 + ], + "ɦ": [ + 71 + ], + "ɧ": [ + 72 + ], + "ɨ": [ + 73 + ], + "ɪ": [ + 74 + ], + "ɫ": [ + 75 + ], + "ɬ": [ + 76 + ], + "ɭ": [ + 77 + ], + "ɮ": [ + 78 + ], + "ɯ": [ + 79 + ], + "ɰ": [ + 80 + ], + "ɱ": [ + 81 + ], + "ɲ": [ + 82 + ], + "ɳ": [ + 83 + ], + "ɴ": [ + 84 + ], + "ɵ": [ + 85 + ], + "ɶ": [ + 86 + ], + "ɸ": [ + 87 + ], + "ɹ": [ + 88 + ], + "ɺ": [ + 89 + ], + "ɻ": [ + 90 + ], + "ɽ": [ + 91 + ], + "ɾ": [ + 92 + ], + "ʀ": [ + 93 + ], + "ʁ": [ + 94 + ], + "ʂ": [ + 95 + ], + "ʃ": [ + 96 + ], + "ʄ": [ + 97 + ], + "ʈ": [ + 98 + ], + "ʉ": [ + 99 + ], + "ʊ": [ + 100 + ], + "ʋ": [ + 101 + ], + "ʌ": [ + 102 + ], + "ʍ": [ + 103 + ], + "ʎ": [ + 104 + ], + "ʏ": [ + 105 + ], + "ʐ": [ + 106 + ], + "ʑ": [ + 107 + ], + "ʒ": [ + 108 + ], + "ʔ": [ + 109 + ], + "ʕ": [ + 110 + ], + "ʘ": [ + 111 + ], + "ʙ": [ + 112 + ], + "ʛ": [ + 113 + ], + "ʜ": [ + 114 + ], + "ʝ": [ + 115 + ], + "ʟ": [ + 116 + ], + "ʡ": [ + 117 + ], + "ʢ": [ + 118 + ], + "ʦ": [ + 155 + ], + "ʰ": [ + 145 + ], + "ʲ": [ + 119 + ], + "ˈ": [ + 120 + ], + "ˌ": [ + 121 + ], + "ː": [ + 122 + ], + "ˑ": [ + 123 + ], + "˞": [ + 124 + ], + "ˤ": [ + 146 + ], + "̃": [ + 141 + ], + "̧": [ + 140 + ], + "̩": [ + 144 + ], + "̪": [ + 142 + ], + "̯": [ + 143 + ], + "̺": [ + 152 + ], + "̻": [ + 153 + ], + "β": [ + 125 + ], + "ε": [ + 147 + ], + "θ": [ + 126 + ], + "χ": [ + 127 + ], + "ᵻ": [ + 128 + ], + "↑": [ + 151 + ], + "↓": [ + 148 + ], + "ⱱ": [ + 129 + ] + }, + "num_symbols": 256, + "num_speakers": 1, + "speaker_id_map": {}, + "language": { + "code": "lb_LU", + "family": "lb", + "region": "LU", + "name_native": "Lëtzebuergesch", + "name_english": "Luxembourgish", + "country_english": "Luxembourg" + }, + "dataset": "marylux" +} \ No newline at end of file diff --git a/lb/lb_LU/marylux/medium/samples/speaker_0.mp3 b/lb/lb_LU/marylux/medium/samples/speaker_0.mp3 new file mode 100644 index 0000000000000000000000000000000000000000..002dd3061ddaa303efcdf49e8501fdd41c5087dd Binary files /dev/null and b/lb/lb_LU/marylux/medium/samples/speaker_0.mp3 differ diff --git a/ne/ne_NP/google/medium/ALIASES b/ne/ne_NP/google/medium/ALIASES new file mode 100644 index 0000000000000000000000000000000000000000..afeda5d596bc17ae7c05906216127a161a521ed9 --- /dev/null +++ b/ne/ne_NP/google/medium/ALIASES @@ -0,0 +1 @@ +ne-google-medium diff --git a/ne/ne_NP/google/medium/MODEL_CARD b/ne/ne_NP/google/medium/MODEL_CARD new file mode 100644 index 0000000000000000000000000000000000000000..e205115b5d390670e56f3a89c4ad84271589ddb7 --- /dev/null +++ b/ne/ne_NP/google/medium/MODEL_CARD @@ -0,0 +1,15 @@ +# Model card for google (medium) + +* Language: ne_NP (Nepali, Nepal) +* Speakers: 18 +* Quality: medium +* Samplerate: 22,050Hz + +## Dataset + +* URL: http://www.openslr.org/43/ +* License: CC-BY-SA-4.0 International + +## Training + +Finetuned from U.S. English lessac voice (medium quality). diff --git a/ne/ne_NP/google/medium/ne_NP-google-medium.onnx b/ne/ne_NP/google/medium/ne_NP-google-medium.onnx new file mode 100644 index 0000000000000000000000000000000000000000..632dfd16e441294c84e2125fdcffdc9775eead5f --- /dev/null +++ b/ne/ne_NP/google/medium/ne_NP-google-medium.onnx @@ -0,0 +1,3 @@ +version https://git-lfs.github.com/spec/v1 +oid sha256:e3ff3cbf97a7c01ebf29263c7fa1899ebed15e27a2d819b93dcfb86e10d39eaa +size 76766385 diff --git a/ne/ne_NP/google/medium/ne_NP-google-medium.onnx.json b/ne/ne_NP/google/medium/ne_NP-google-medium.onnx.json new file mode 100644 index 0000000000000000000000000000000000000000..ce252c85e5e13818247503ab9e4ed9a675cc2fa1 --- /dev/null +++ b/ne/ne_NP/google/medium/ne_NP-google-medium.onnx.json @@ -0,0 +1,512 @@ +{ + "audio": { + "sample_rate": 22050, + "quality": "medium" + }, + "espeak": { + "voice": "ne" + }, + "inference": { + "noise_scale": 0.667, + "length_scale": 1, + "noise_w": 0.8 + }, + "phoneme_type": "espeak", + "phoneme_map": {}, + "phoneme_id_map": { + "_": [ + 0 + ], + "^": [ + 1 + ], + "$": [ + 2 + ], + " ": [ + 3 + ], + "!": [ + 4 + ], + "'": [ + 5 + ], + "(": [ + 6 + ], + ")": [ + 7 + ], + ",": [ + 8 + ], + "-": [ + 9 + ], + ".": [ + 10 + ], + ":": [ + 11 + ], + ";": [ + 12 + ], + "?": [ + 13 + ], + "a": [ + 14 + ], + "b": [ + 15 + ], + "c": [ + 16 + ], + "d": [ + 17 + ], + "e": [ + 18 + ], + "f": [ + 19 + ], + "h": [ + 20 + ], + "i": [ + 21 + ], + "j": [ + 22 + ], + "k": [ + 23 + ], + "l": [ + 24 + ], + "m": [ + 25 + ], + "n": [ + 26 + ], + "o": [ + 27 + ], + "p": [ + 28 + ], + "q": [ + 29 + ], + "r": [ + 30 + ], + "s": [ + 31 + ], + "t": [ + 32 + ], + "u": [ + 33 + ], + "v": [ + 34 + ], + "w": [ + 35 + ], + "x": [ + 36 + ], + "y": [ + 37 + ], + "z": [ + 38 + ], + "æ": [ + 39 + ], + "ç": [ + 40 + ], + "ð": [ + 41 + ], + "ø": [ + 42 + ], + "ħ": [ + 43 + ], + "ŋ": [ + 44 + ], + "œ": [ + 45 + ], + "ǀ": [ + 46 + ], + "ǁ": [ + 47 + ], + "ǂ": [ + 48 + ], + "ǃ": [ + 49 + ], + "ɐ": [ + 50 + ], + "ɑ": [ + 51 + ], + "ɒ": [ + 52 + ], + "ɓ": [ + 53 + ], + "ɔ": [ + 54 + ], + "ɕ": [ + 55 + ], + "ɖ": [ + 56 + ], + "ɗ": [ + 57 + ], + "ɘ": [ + 58 + ], + "ə": [ + 59 + ], + "ɚ": [ + 60 + ], + "ɛ": [ + 61 + ], + "ɜ": [ + 62 + ], + "ɞ": [ + 63 + ], + "ɟ": [ + 64 + ], + "ɠ": [ + 65 + ], + "ɡ": [ + 66 + ], + "ɢ": [ + 67 + ], + "ɣ": [ + 68 + ], + "ɤ": [ + 69 + ], + "ɥ": [ + 70 + ], + "ɦ": [ + 71 + ], + "ɧ": [ + 72 + ], + "ɨ": [ + 73 + ], + "ɪ": [ + 74 + ], + "ɫ": [ + 75 + ], + "ɬ": [ + 76 + ], + "ɭ": [ + 77 + ], + "ɮ": [ + 78 + ], + "ɯ": [ + 79 + ], + "ɰ": [ + 80 + ], + "ɱ": [ + 81 + ], + "ɲ": [ + 82 + ], + "ɳ": [ + 83 + ], + "ɴ": [ + 84 + ], + "ɵ": [ + 85 + ], + "ɶ": [ + 86 + ], + "ɸ": [ + 87 + ], + "ɹ": [ + 88 + ], + "ɺ": [ + 89 + ], + "ɻ": [ + 90 + ], + "ɽ": [ + 91 + ], + "ɾ": [ + 92 + ], + "ʀ": [ + 93 + ], + "ʁ": [ + 94 + ], + "ʂ": [ + 95 + ], + "ʃ": [ + 96 + ], + "ʄ": [ + 97 + ], + "ʈ": [ + 98 + ], + "ʉ": [ + 99 + ], + "ʊ": [ + 100 + ], + "ʋ": [ + 101 + ], + "ʌ": [ + 102 + ], + "ʍ": [ + 103 + ], + "ʎ": [ + 104 + ], + "ʏ": [ + 105 + ], + "ʐ": [ + 106 + ], + "ʑ": [ + 107 + ], + "ʒ": [ + 108 + ], + "ʔ": [ + 109 + ], + "ʕ": [ + 110 + ], + "ʘ": [ + 111 + ], + "ʙ": [ + 112 + ], + "ʛ": [ + 113 + ], + "ʜ": [ + 114 + ], + "ʝ": [ + 115 + ], + "ʟ": [ + 116 + ], + "ʡ": [ + 117 + ], + "ʢ": [ + 118 + ], + "ʲ": [ + 119 + ], + "ˈ": [ + 120 + ], + "ˌ": [ + 121 + ], + "ː": [ + 122 + ], + "ˑ": [ + 123 + ], + "˞": [ + 124 + ], + "β": [ + 125 + ], + "θ": [ + 126 + ], + "χ": [ + 127 + ], + "ᵻ": [ + 128 + ], + "ⱱ": [ + 129 + ], + "0": [ + 130 + ], + "1": [ + 131 + ], + "2": [ + 132 + ], + "3": [ + 133 + ], + "4": [ + 134 + ], + "5": [ + 135 + ], + "6": [ + 136 + ], + "7": [ + 137 + ], + "8": [ + 138 + ], + "9": [ + 139 + ], + "̧": [ + 140 + ], + "̃": [ + 141 + ], + "̪": [ + 142 + ], + "̯": [ + 143 + ], + "̩": [ + 144 + ], + "ʰ": [ + 145 + ], + "ˤ": [ + 146 + ], + "ε": [ + 147 + ], + "↓": [ + 148 + ], + "#": [ + 149 + ], + "\"": [ + 150 + ], + "↑": [ + 151 + ], + "̺": [ + 152 + ], + "̻": [ + 153 + ] + }, + "num_symbols": 256, + "num_speakers": 18, + "speaker_id_map": { + "0546": 0, + "3614": 1, + "2099": 2, + "3960": 3, + "6834": 4, + "7957": 5, + "6329": 6, + "9407": 7, + "6587": 8, + "0258": 9, + "2139": 10, + "5687": 11, + "0283": 12, + "3997": 13, + "3154": 14, + "0883": 15, + "2027": 16, + "0649": 17 + }, + "piper_version": "1.0.0", + "language": { + "code": "ne_NP", + "family": "ne", + "region": "NP", + "name_native": "नेपाली", + "name_english": "Nepali", + "country_english": "Nepal" + }, + "dataset": "google" +} \ No newline at end of file diff --git a/ne/ne_NP/google/medium/samples/speaker_0.mp3 b/ne/ne_NP/google/medium/samples/speaker_0.mp3 new file mode 100644 index 0000000000000000000000000000000000000000..02cac8075f29c941e7e571ea792c51873b30f843 Binary files /dev/null and b/ne/ne_NP/google/medium/samples/speaker_0.mp3 differ diff --git a/ne/ne_NP/google/medium/samples/speaker_1.mp3 b/ne/ne_NP/google/medium/samples/speaker_1.mp3 new file mode 100644 index 0000000000000000000000000000000000000000..4081e76a6c24f54e8e9054c33fdfd07035a25a59 Binary files /dev/null and b/ne/ne_NP/google/medium/samples/speaker_1.mp3 differ diff --git a/ne/ne_NP/google/medium/samples/speaker_10.mp3 b/ne/ne_NP/google/medium/samples/speaker_10.mp3 new file mode 100644 index 0000000000000000000000000000000000000000..32f946d9b0b5fe5f20010fc2fb66fbb4bba704bb Binary files /dev/null and b/ne/ne_NP/google/medium/samples/speaker_10.mp3 differ diff --git a/ne/ne_NP/google/medium/samples/speaker_11.mp3 b/ne/ne_NP/google/medium/samples/speaker_11.mp3 new file mode 100644 index 0000000000000000000000000000000000000000..2db0c3de4a1d9f763ec6cbec09f2e251b0b90bba Binary files /dev/null and b/ne/ne_NP/google/medium/samples/speaker_11.mp3 differ diff --git a/ne/ne_NP/google/medium/samples/speaker_12.mp3 b/ne/ne_NP/google/medium/samples/speaker_12.mp3 new file mode 100644 index 0000000000000000000000000000000000000000..7426bd7c90bcf9377fb3299ca6551878cae5aa07 Binary files /dev/null and b/ne/ne_NP/google/medium/samples/speaker_12.mp3 differ diff --git a/ne/ne_NP/google/medium/samples/speaker_13.mp3 b/ne/ne_NP/google/medium/samples/speaker_13.mp3 new file mode 100644 index 0000000000000000000000000000000000000000..8f7908559fd141da8814e1cc7ee10c1d72537cce Binary files /dev/null and b/ne/ne_NP/google/medium/samples/speaker_13.mp3 differ diff --git a/ne/ne_NP/google/medium/samples/speaker_14.mp3 b/ne/ne_NP/google/medium/samples/speaker_14.mp3 new file mode 100644 index 0000000000000000000000000000000000000000..858fb0e1dcc6158f154917da551359543118dc94 Binary files /dev/null and b/ne/ne_NP/google/medium/samples/speaker_14.mp3 differ diff --git a/ne/ne_NP/google/medium/samples/speaker_15.mp3 b/ne/ne_NP/google/medium/samples/speaker_15.mp3 new file mode 100644 index 0000000000000000000000000000000000000000..f5991192e15cbb7bf2a5daf0eab1e3e245a67ba2 Binary files /dev/null and b/ne/ne_NP/google/medium/samples/speaker_15.mp3 differ diff --git a/ne/ne_NP/google/medium/samples/speaker_16.mp3 b/ne/ne_NP/google/medium/samples/speaker_16.mp3 new file mode 100644 index 0000000000000000000000000000000000000000..b6bc1cc018bde68bfcd0518cc636062a4fd2ac2d Binary files /dev/null and b/ne/ne_NP/google/medium/samples/speaker_16.mp3 differ diff --git a/ne/ne_NP/google/medium/samples/speaker_17.mp3 b/ne/ne_NP/google/medium/samples/speaker_17.mp3 new file mode 100644 index 0000000000000000000000000000000000000000..21cd15a4e93ce627531c39373244f68c813cf611 Binary files /dev/null and b/ne/ne_NP/google/medium/samples/speaker_17.mp3 differ diff --git a/ne/ne_NP/google/medium/samples/speaker_2.mp3 b/ne/ne_NP/google/medium/samples/speaker_2.mp3 new file mode 100644 index 0000000000000000000000000000000000000000..f55607dbee65f459c8c8a9c148d6c3da0e61b5fc Binary files /dev/null and b/ne/ne_NP/google/medium/samples/speaker_2.mp3 differ diff --git a/ne/ne_NP/google/medium/samples/speaker_3.mp3 b/ne/ne_NP/google/medium/samples/speaker_3.mp3 new file mode 100644 index 0000000000000000000000000000000000000000..b7d2157ddc3334332fd7672cbcacd3723fe30e49 Binary files /dev/null and b/ne/ne_NP/google/medium/samples/speaker_3.mp3 differ diff --git a/ne/ne_NP/google/medium/samples/speaker_4.mp3 b/ne/ne_NP/google/medium/samples/speaker_4.mp3 new file mode 100644 index 0000000000000000000000000000000000000000..9250564790da41885fdd4e4ac301410d749c39f4 Binary files /dev/null and b/ne/ne_NP/google/medium/samples/speaker_4.mp3 differ diff --git a/ne/ne_NP/google/medium/samples/speaker_5.mp3 b/ne/ne_NP/google/medium/samples/speaker_5.mp3 new file mode 100644 index 0000000000000000000000000000000000000000..e1a3808d213a8cb37bfbfb712812dbfde9c110af Binary files /dev/null and b/ne/ne_NP/google/medium/samples/speaker_5.mp3 differ diff --git a/ne/ne_NP/google/medium/samples/speaker_6.mp3 b/ne/ne_NP/google/medium/samples/speaker_6.mp3 new file mode 100644 index 0000000000000000000000000000000000000000..be636f163e0675b4632657c16c4248fd355b597f Binary files /dev/null and b/ne/ne_NP/google/medium/samples/speaker_6.mp3 differ diff --git a/ne/ne_NP/google/medium/samples/speaker_7.mp3 b/ne/ne_NP/google/medium/samples/speaker_7.mp3 new file mode 100644 index 0000000000000000000000000000000000000000..b78ba6055441e721196e493cbb32df96ecf45fe3 Binary files /dev/null and b/ne/ne_NP/google/medium/samples/speaker_7.mp3 differ diff --git a/ne/ne_NP/google/medium/samples/speaker_8.mp3 b/ne/ne_NP/google/medium/samples/speaker_8.mp3 new file mode 100644 index 0000000000000000000000000000000000000000..3d00e3f38d9f19fc62b181bc27a5a83fe43200fa Binary files /dev/null and b/ne/ne_NP/google/medium/samples/speaker_8.mp3 differ diff --git a/ne/ne_NP/google/medium/samples/speaker_9.mp3 b/ne/ne_NP/google/medium/samples/speaker_9.mp3 new file mode 100644 index 0000000000000000000000000000000000000000..9b24907ec9f9fc71ec4816cdcd25920fbdb4fe25 Binary files /dev/null and b/ne/ne_NP/google/medium/samples/speaker_9.mp3 differ diff --git a/ne/ne_NP/google/x_low/ALIASES b/ne/ne_NP/google/x_low/ALIASES new file mode 100644 index 0000000000000000000000000000000000000000..d272f7596cca716b3b18c9115afd806835441bc3 --- /dev/null +++ b/ne/ne_NP/google/x_low/ALIASES @@ -0,0 +1 @@ +ne-google-x-low diff --git a/ne/ne_NP/google/x_low/MODEL_CARD b/ne/ne_NP/google/x_low/MODEL_CARD new file mode 100644 index 0000000000000000000000000000000000000000..ffc4c3729bfc992c0c5c6037cc0d6d11e8b0654f --- /dev/null +++ b/ne/ne_NP/google/x_low/MODEL_CARD @@ -0,0 +1,15 @@ +# Model card for google (x_low) + +* Language: ne_NE (Nepali, Nepal) +* Speakers: 18 +* Quality: x_low +* Samplerate: 16,000Hz + +## Dataset + +* URL: http://www.openslr.org/43/ +* License: CC-BY-SA-4.0 International + +## Training + +Trained from scratch. diff --git a/ne/ne_NP/google/x_low/ne_NP-google-x_low.onnx b/ne/ne_NP/google/x_low/ne_NP-google-x_low.onnx new file mode 100644 index 0000000000000000000000000000000000000000..577ee91e2a7b8c23b5acc6c277b5ced2f27efe1c --- /dev/null +++ b/ne/ne_NP/google/x_low/ne_NP-google-x_low.onnx @@ -0,0 +1,3 @@ +version https://git-lfs.github.com/spec/v1 +oid sha256:283431c1f89be9b13a40f885da818b23ac13f6ca536e2a0a414c970d3b13f0f2 +size 27693157 diff --git a/ne/ne_NP/google/x_low/ne_NP-google-x_low.onnx.json b/ne/ne_NP/google/x_low/ne_NP-google-x_low.onnx.json new file mode 100644 index 0000000000000000000000000000000000000000..7d141b1da4640bfa99207b4795be98814e512d4d --- /dev/null +++ b/ne/ne_NP/google/x_low/ne_NP-google-x_low.onnx.json @@ -0,0 +1,439 @@ +{ + "audio": { + "sample_rate": 16000, + "quality": "x_low" + }, + "espeak": { + "voice": "ne" + }, + "inference": { + "noise_scale": 0.667, + "length_scale": 1, + "noise_w": 0.8 + }, + "phoneme_map": {}, + "phoneme_id_map": { + "_": [ + 0 + ], + "^": [ + 1 + ], + "$": [ + 2 + ], + " ": [ + 3 + ], + "!": [ + 4 + ], + "'": [ + 5 + ], + "(": [ + 6 + ], + ")": [ + 7 + ], + ",": [ + 8 + ], + "-": [ + 9 + ], + ".": [ + 10 + ], + ":": [ + 11 + ], + ";": [ + 12 + ], + "?": [ + 13 + ], + "a": [ + 14 + ], + "b": [ + 15 + ], + "c": [ + 16 + ], + "d": [ + 17 + ], + "e": [ + 18 + ], + "f": [ + 19 + ], + "h": [ + 20 + ], + "i": [ + 21 + ], + "j": [ + 22 + ], + "k": [ + 23 + ], + "l": [ + 24 + ], + "m": [ + 25 + ], + "n": [ + 26 + ], + "o": [ + 27 + ], + "p": [ + 28 + ], + "q": [ + 29 + ], + "r": [ + 30 + ], + "s": [ + 31 + ], + "t": [ + 32 + ], + "u": [ + 33 + ], + "v": [ + 34 + ], + "w": [ + 35 + ], + "x": [ + 36 + ], + "y": [ + 37 + ], + "z": [ + 38 + ], + "æ": [ + 39 + ], + "ç": [ + 40 + ], + "ð": [ + 41 + ], + "ø": [ + 42 + ], + "ħ": [ + 43 + ], + "ŋ": [ + 44 + ], + "œ": [ + 45 + ], + "ǀ": [ + 46 + ], + "ǁ": [ + 47 + ], + "ǂ": [ + 48 + ], + "ǃ": [ + 49 + ], + "ɐ": [ + 50 + ], + "ɑ": [ + 51 + ], + "ɒ": [ + 52 + ], + "ɓ": [ + 53 + ], + "ɔ": [ + 54 + ], + "ɕ": [ + 55 + ], + "ɖ": [ + 56 + ], + "ɗ": [ + 57 + ], + "ɘ": [ + 58 + ], + "ə": [ + 59 + ], + "ɚ": [ + 60 + ], + "ɛ": [ + 61 + ], + "ɜ": [ + 62 + ], + "ɞ": [ + 63 + ], + "ɟ": [ + 64 + ], + "ɠ": [ + 65 + ], + "ɡ": [ + 66 + ], + "ɢ": [ + 67 + ], + "ɣ": [ + 68 + ], + "ɤ": [ + 69 + ], + "ɥ": [ + 70 + ], + "ɦ": [ + 71 + ], + "ɧ": [ + 72 + ], + "ɨ": [ + 73 + ], + "ɪ": [ + 74 + ], + "ɫ": [ + 75 + ], + "ɬ": [ + 76 + ], + "ɭ": [ + 77 + ], + "ɮ": [ + 78 + ], + "ɯ": [ + 79 + ], + "ɰ": [ + 80 + ], + "ɱ": [ + 81 + ], + "ɲ": [ + 82 + ], + "ɳ": [ + 83 + ], + "ɴ": [ + 84 + ], + "ɵ": [ + 85 + ], + "ɶ": [ + 86 + ], + "ɸ": [ + 87 + ], + "ɹ": [ + 88 + ], + "ɺ": [ + 89 + ], + "ɻ": [ + 90 + ], + "ɽ": [ + 91 + ], + "ɾ": [ + 92 + ], + "ʀ": [ + 93 + ], + "ʁ": [ + 94 + ], + "ʂ": [ + 95 + ], + "ʃ": [ + 96 + ], + "ʄ": [ + 97 + ], + "ʈ": [ + 98 + ], + "ʉ": [ + 99 + ], + "ʊ": [ + 100 + ], + "ʋ": [ + 101 + ], + "ʌ": [ + 102 + ], + "ʍ": [ + 103 + ], + "ʎ": [ + 104 + ], + "ʏ": [ + 105 + ], + "ʐ": [ + 106 + ], + "ʑ": [ + 107 + ], + "ʒ": [ + 108 + ], + "ʔ": [ + 109 + ], + "ʕ": [ + 110 + ], + "ʘ": [ + 111 + ], + "ʙ": [ + 112 + ], + "ʛ": [ + 113 + ], + "ʜ": [ + 114 + ], + "ʝ": [ + 115 + ], + "ʟ": [ + 116 + ], + "ʡ": [ + 117 + ], + "ʢ": [ + 118 + ], + "ʲ": [ + 119 + ], + "ˈ": [ + 120 + ], + "ˌ": [ + 121 + ], + "ː": [ + 122 + ], + "ˑ": [ + 123 + ], + "˞": [ + 124 + ], + "β": [ + 125 + ], + "θ": [ + 126 + ], + "χ": [ + 127 + ], + "ᵻ": [ + 128 + ], + "ⱱ": [ + 129 + ] + }, + "num_symbols": 130, + "num_speakers": 18, + "speaker_id_map": { + "0546": 0, + "3614": 1, + "2099": 2, + "3960": 3, + "6834": 4, + "7957": 5, + "6329": 6, + "9407": 7, + "6587": 8, + "0258": 9, + "2139": 10, + "5687": 11, + "0283": 12, + "3997": 13, + "3154": 14, + "0883": 15, + "2027": 16, + "0649": 17 + }, + "piper_version": "0.2.0", + "language": { + "code": "ne_NP", + "family": "ne", + "region": "NP", + "name_native": "नेपाली", + "name_english": "Nepali", + "country_english": "Nepal" + }, + "dataset": "google" +} \ No newline at end of file diff --git a/ne/ne_NP/google/x_low/samples/speaker_0.mp3 b/ne/ne_NP/google/x_low/samples/speaker_0.mp3 new file mode 100644 index 0000000000000000000000000000000000000000..3e85e578d695a28797bd573be9a607c1c56d4d48 Binary files /dev/null and b/ne/ne_NP/google/x_low/samples/speaker_0.mp3 differ diff --git a/ne/ne_NP/google/x_low/samples/speaker_1.mp3 b/ne/ne_NP/google/x_low/samples/speaker_1.mp3 new file mode 100644 index 0000000000000000000000000000000000000000..2f1bfd5250a17e938b52a336797182bc542f2c9d Binary files /dev/null and b/ne/ne_NP/google/x_low/samples/speaker_1.mp3 differ diff --git a/ne/ne_NP/google/x_low/samples/speaker_10.mp3 b/ne/ne_NP/google/x_low/samples/speaker_10.mp3 new file mode 100644 index 0000000000000000000000000000000000000000..956262301fc953646c3f5aa6dedb4428a976ff70 Binary files /dev/null and b/ne/ne_NP/google/x_low/samples/speaker_10.mp3 differ diff --git a/ne/ne_NP/google/x_low/samples/speaker_11.mp3 b/ne/ne_NP/google/x_low/samples/speaker_11.mp3 new file mode 100644 index 0000000000000000000000000000000000000000..172683243f9a86969ef79ecec321fc2733e7aa24 Binary files /dev/null and b/ne/ne_NP/google/x_low/samples/speaker_11.mp3 differ diff --git a/ne/ne_NP/google/x_low/samples/speaker_12.mp3 b/ne/ne_NP/google/x_low/samples/speaker_12.mp3 new file mode 100644 index 0000000000000000000000000000000000000000..779180941a265b71bb9b786a92b21e8a3da66cfa Binary files /dev/null and b/ne/ne_NP/google/x_low/samples/speaker_12.mp3 differ diff --git a/ne/ne_NP/google/x_low/samples/speaker_13.mp3 b/ne/ne_NP/google/x_low/samples/speaker_13.mp3 new file mode 100644 index 0000000000000000000000000000000000000000..f0c8b1e615d952731f18c3cc03c4b85a6eb4c964 Binary files /dev/null and b/ne/ne_NP/google/x_low/samples/speaker_13.mp3 differ diff --git a/ne/ne_NP/google/x_low/samples/speaker_14.mp3 b/ne/ne_NP/google/x_low/samples/speaker_14.mp3 new file mode 100644 index 0000000000000000000000000000000000000000..a4c753455b382e1a85363a6149d5c6fec89f71e6 Binary files /dev/null and b/ne/ne_NP/google/x_low/samples/speaker_14.mp3 differ diff --git a/ne/ne_NP/google/x_low/samples/speaker_15.mp3 b/ne/ne_NP/google/x_low/samples/speaker_15.mp3 new file mode 100644 index 0000000000000000000000000000000000000000..e1798af58984c2075dd76fa1a60ca3bb21ad8453 Binary files /dev/null and b/ne/ne_NP/google/x_low/samples/speaker_15.mp3 differ diff --git a/ne/ne_NP/google/x_low/samples/speaker_16.mp3 b/ne/ne_NP/google/x_low/samples/speaker_16.mp3 new file mode 100644 index 0000000000000000000000000000000000000000..8c720e277513ea4d51064cd93a20421372430072 Binary files /dev/null and b/ne/ne_NP/google/x_low/samples/speaker_16.mp3 differ diff --git a/ne/ne_NP/google/x_low/samples/speaker_17.mp3 b/ne/ne_NP/google/x_low/samples/speaker_17.mp3 new file mode 100644 index 0000000000000000000000000000000000000000..3997491bd14ce4a550ebbcd0aa045d8f3a720235 Binary files /dev/null and b/ne/ne_NP/google/x_low/samples/speaker_17.mp3 differ diff --git a/ne/ne_NP/google/x_low/samples/speaker_2.mp3 b/ne/ne_NP/google/x_low/samples/speaker_2.mp3 new file mode 100644 index 0000000000000000000000000000000000000000..0d964454d4b2a776dcc4616dc6028fd1e8ca60ea Binary files /dev/null and b/ne/ne_NP/google/x_low/samples/speaker_2.mp3 differ diff --git a/ne/ne_NP/google/x_low/samples/speaker_3.mp3 b/ne/ne_NP/google/x_low/samples/speaker_3.mp3 new file mode 100644 index 0000000000000000000000000000000000000000..e68b3d9db3b9363360f9b1a089c0941898183129 Binary files /dev/null and b/ne/ne_NP/google/x_low/samples/speaker_3.mp3 differ diff --git a/ne/ne_NP/google/x_low/samples/speaker_4.mp3 b/ne/ne_NP/google/x_low/samples/speaker_4.mp3 new file mode 100644 index 0000000000000000000000000000000000000000..9bec8894538f60a8e9bd56e690c06ad0b4da61f5 Binary files /dev/null and b/ne/ne_NP/google/x_low/samples/speaker_4.mp3 differ diff --git a/ne/ne_NP/google/x_low/samples/speaker_5.mp3 b/ne/ne_NP/google/x_low/samples/speaker_5.mp3 new file mode 100644 index 0000000000000000000000000000000000000000..09334209aa94d7feef5a0554e653351382c093d2 Binary files /dev/null and b/ne/ne_NP/google/x_low/samples/speaker_5.mp3 differ diff --git a/ne/ne_NP/google/x_low/samples/speaker_6.mp3 b/ne/ne_NP/google/x_low/samples/speaker_6.mp3 new file mode 100644 index 0000000000000000000000000000000000000000..47b6458266880b5b8d09d7648ac08c5dc9c5f9f0 Binary files /dev/null and b/ne/ne_NP/google/x_low/samples/speaker_6.mp3 differ diff --git a/ne/ne_NP/google/x_low/samples/speaker_7.mp3 b/ne/ne_NP/google/x_low/samples/speaker_7.mp3 new file mode 100644 index 0000000000000000000000000000000000000000..968cdac77b791fad35197b275b0a3705be409d78 Binary files /dev/null and b/ne/ne_NP/google/x_low/samples/speaker_7.mp3 differ diff --git a/ne/ne_NP/google/x_low/samples/speaker_8.mp3 b/ne/ne_NP/google/x_low/samples/speaker_8.mp3 new file mode 100644 index 0000000000000000000000000000000000000000..18ef171246579028a373d9c082dd4dfec5adff78 Binary files /dev/null and b/ne/ne_NP/google/x_low/samples/speaker_8.mp3 differ diff --git a/ne/ne_NP/google/x_low/samples/speaker_9.mp3 b/ne/ne_NP/google/x_low/samples/speaker_9.mp3 new file mode 100644 index 0000000000000000000000000000000000000000..2baac7ea155828413da20ee65afbb22867312731 Binary files /dev/null and b/ne/ne_NP/google/x_low/samples/speaker_9.mp3 differ diff --git a/nl/nl_BE/nathalie/medium/MODEL_CARD b/nl/nl_BE/nathalie/medium/MODEL_CARD new file mode 100644 index 0000000000000000000000000000000000000000..9874ec439afb104991f1e4c545d163c93744d267 --- /dev/null +++ b/nl/nl_BE/nathalie/medium/MODEL_CARD @@ -0,0 +1,15 @@ +# Model card for nathalie (medium) + +* Language: nl_BE (Dutch, Belgium) +* Speakers: 1 +* Quality: medium +* Samplerate: 22,050Hz + +## Dataset + +* URL: https://github.com/rhasspy/dataset-voice-nathalie +* License: CC0 + +## Training + +Finetuned from U.S. English lessac voice (medium quality) diff --git a/nl/nl_BE/nathalie/medium/nl_BE-nathalie-medium.onnx b/nl/nl_BE/nathalie/medium/nl_BE-nathalie-medium.onnx new file mode 100644 index 0000000000000000000000000000000000000000..ce24583163ae7d0a9b07d3f40746cf26c638f343 --- /dev/null +++ b/nl/nl_BE/nathalie/medium/nl_BE-nathalie-medium.onnx @@ -0,0 +1,3 @@ +version https://git-lfs.github.com/spec/v1 +oid sha256:49cf48023861f9fd42e13a8632f068fee67d1ce244a6ee38f29595afbf0a6be4 +size 63201294 diff --git a/nl/nl_BE/nathalie/medium/nl_BE-nathalie-medium.onnx.json b/nl/nl_BE/nathalie/medium/nl_BE-nathalie-medium.onnx.json new file mode 100644 index 0000000000000000000000000000000000000000..896fe1868b89a986d6d647df06988eb92f337088 --- /dev/null +++ b/nl/nl_BE/nathalie/medium/nl_BE-nathalie-medium.onnx.json @@ -0,0 +1,493 @@ +{ + "audio": { + "sample_rate": 22050, + "quality": "medium" + }, + "espeak": { + "voice": "nl" + }, + "inference": { + "noise_scale": 0.667, + "length_scale": 1, + "noise_w": 0.8 + }, + "phoneme_type": "espeak", + "phoneme_map": {}, + "phoneme_id_map": { + "_": [ + 0 + ], + "^": [ + 1 + ], + "$": [ + 2 + ], + " ": [ + 3 + ], + "!": [ + 4 + ], + "'": [ + 5 + ], + "(": [ + 6 + ], + ")": [ + 7 + ], + ",": [ + 8 + ], + "-": [ + 9 + ], + ".": [ + 10 + ], + ":": [ + 11 + ], + ";": [ + 12 + ], + "?": [ + 13 + ], + "a": [ + 14 + ], + "b": [ + 15 + ], + "c": [ + 16 + ], + "d": [ + 17 + ], + "e": [ + 18 + ], + "f": [ + 19 + ], + "h": [ + 20 + ], + "i": [ + 21 + ], + "j": [ + 22 + ], + "k": [ + 23 + ], + "l": [ + 24 + ], + "m": [ + 25 + ], + "n": [ + 26 + ], + "o": [ + 27 + ], + "p": [ + 28 + ], + "q": [ + 29 + ], + "r": [ + 30 + ], + "s": [ + 31 + ], + "t": [ + 32 + ], + "u": [ + 33 + ], + "v": [ + 34 + ], + "w": [ + 35 + ], + "x": [ + 36 + ], + "y": [ + 37 + ], + "z": [ + 38 + ], + "æ": [ + 39 + ], + "ç": [ + 40 + ], + "ð": [ + 41 + ], + "ø": [ + 42 + ], + "ħ": [ + 43 + ], + "ŋ": [ + 44 + ], + "œ": [ + 45 + ], + "ǀ": [ + 46 + ], + "ǁ": [ + 47 + ], + "ǂ": [ + 48 + ], + "ǃ": [ + 49 + ], + "ɐ": [ + 50 + ], + "ɑ": [ + 51 + ], + "ɒ": [ + 52 + ], + "ɓ": [ + 53 + ], + "ɔ": [ + 54 + ], + "ɕ": [ + 55 + ], + "ɖ": [ + 56 + ], + "ɗ": [ + 57 + ], + "ɘ": [ + 58 + ], + "ə": [ + 59 + ], + "ɚ": [ + 60 + ], + "ɛ": [ + 61 + ], + "ɜ": [ + 62 + ], + "ɞ": [ + 63 + ], + "ɟ": [ + 64 + ], + "ɠ": [ + 65 + ], + "ɡ": [ + 66 + ], + "ɢ": [ + 67 + ], + "ɣ": [ + 68 + ], + "ɤ": [ + 69 + ], + "ɥ": [ + 70 + ], + "ɦ": [ + 71 + ], + "ɧ": [ + 72 + ], + "ɨ": [ + 73 + ], + "ɪ": [ + 74 + ], + "ɫ": [ + 75 + ], + "ɬ": [ + 76 + ], + "ɭ": [ + 77 + ], + "ɮ": [ + 78 + ], + "ɯ": [ + 79 + ], + "ɰ": [ + 80 + ], + "ɱ": [ + 81 + ], + "ɲ": [ + 82 + ], + "ɳ": [ + 83 + ], + "ɴ": [ + 84 + ], + "ɵ": [ + 85 + ], + "ɶ": [ + 86 + ], + "ɸ": [ + 87 + ], + "ɹ": [ + 88 + ], + "ɺ": [ + 89 + ], + "ɻ": [ + 90 + ], + "ɽ": [ + 91 + ], + "ɾ": [ + 92 + ], + "ʀ": [ + 93 + ], + "ʁ": [ + 94 + ], + "ʂ": [ + 95 + ], + "ʃ": [ + 96 + ], + "ʄ": [ + 97 + ], + "ʈ": [ + 98 + ], + "ʉ": [ + 99 + ], + "ʊ": [ + 100 + ], + "ʋ": [ + 101 + ], + "ʌ": [ + 102 + ], + "ʍ": [ + 103 + ], + "ʎ": [ + 104 + ], + "ʏ": [ + 105 + ], + "ʐ": [ + 106 + ], + "ʑ": [ + 107 + ], + "ʒ": [ + 108 + ], + "ʔ": [ + 109 + ], + "ʕ": [ + 110 + ], + "ʘ": [ + 111 + ], + "ʙ": [ + 112 + ], + "ʛ": [ + 113 + ], + "ʜ": [ + 114 + ], + "ʝ": [ + 115 + ], + "ʟ": [ + 116 + ], + "ʡ": [ + 117 + ], + "ʢ": [ + 118 + ], + "ʲ": [ + 119 + ], + "ˈ": [ + 120 + ], + "ˌ": [ + 121 + ], + "ː": [ + 122 + ], + "ˑ": [ + 123 + ], + "˞": [ + 124 + ], + "β": [ + 125 + ], + "θ": [ + 126 + ], + "χ": [ + 127 + ], + "ᵻ": [ + 128 + ], + "ⱱ": [ + 129 + ], + "0": [ + 130 + ], + "1": [ + 131 + ], + "2": [ + 132 + ], + "3": [ + 133 + ], + "4": [ + 134 + ], + "5": [ + 135 + ], + "6": [ + 136 + ], + "7": [ + 137 + ], + "8": [ + 138 + ], + "9": [ + 139 + ], + "̧": [ + 140 + ], + "̃": [ + 141 + ], + "̪": [ + 142 + ], + "̯": [ + 143 + ], + "̩": [ + 144 + ], + "ʰ": [ + 145 + ], + "ˤ": [ + 146 + ], + "ε": [ + 147 + ], + "↓": [ + 148 + ], + "#": [ + 149 + ], + "\"": [ + 150 + ], + "↑": [ + 151 + ], + "̺": [ + 152 + ], + "̻": [ + 153 + ] + }, + "num_symbols": 256, + "num_speakers": 1, + "speaker_id_map": {}, + "piper_version": "1.0.0", + "language": { + "code": "nl_BE", + "family": "nl", + "region": "BE", + "name_native": "Nederlands", + "name_english": "Dutch", + "country_english": "Belgium" + }, + "dataset": "nathalie" +} \ No newline at end of file diff --git a/nl/nl_BE/nathalie/medium/samples/speaker_0.mp3 b/nl/nl_BE/nathalie/medium/samples/speaker_0.mp3 new file mode 100644 index 0000000000000000000000000000000000000000..48926b9337f6567caf165dd9c26cce6f7e0494c5 Binary files /dev/null and b/nl/nl_BE/nathalie/medium/samples/speaker_0.mp3 differ diff --git a/nl/nl_BE/nathalie/x_low/ALIASES b/nl/nl_BE/nathalie/x_low/ALIASES new file mode 100644 index 0000000000000000000000000000000000000000..2a67dc56c8015559554c4a1297be5fd73684bef4 --- /dev/null +++ b/nl/nl_BE/nathalie/x_low/ALIASES @@ -0,0 +1 @@ +nl-nathalie-x-low diff --git a/nl/nl_BE/nathalie/x_low/MODEL_CARD b/nl/nl_BE/nathalie/x_low/MODEL_CARD new file mode 100644 index 0000000000000000000000000000000000000000..d19063e8d6ae1451d4d4836b7cba079b1e0af582 --- /dev/null +++ b/nl/nl_BE/nathalie/x_low/MODEL_CARD @@ -0,0 +1,15 @@ +# Model card for nathalie (x_low) + +* Language: nl_BE (Dutch, Belgium) +* Speakers: 1 +* Quality: x_low +* Samplerate: 16,000Hz + +## Dataset + +* URL: https://github.com/rhasspy/dataset-voice-nathalie +* License: CC0 + +## Training + +Trained from scratch. diff --git a/nl/nl_BE/nathalie/x_low/nl_BE-nathalie-x_low.onnx b/nl/nl_BE/nathalie/x_low/nl_BE-nathalie-x_low.onnx new file mode 100644 index 0000000000000000000000000000000000000000..3ba4e1d212c583fe9575cf7d7ce1bd7b94c05c61 --- /dev/null +++ b/nl/nl_BE/nathalie/x_low/nl_BE-nathalie-x_low.onnx @@ -0,0 +1,3 @@ +version https://git-lfs.github.com/spec/v1 +oid sha256:7d005d2d9b3db4326fc4a1ac86b61ab40252ccf97b76d527f4ffd76d4e3f7dbe +size 20628813 diff --git a/nl/nl_BE/nathalie/x_low/nl_BE-nathalie-x_low.onnx.json b/nl/nl_BE/nathalie/x_low/nl_BE-nathalie-x_low.onnx.json new file mode 100644 index 0000000000000000000000000000000000000000..cf5d7db197f919d5fbf00cb481770aea68ac9b54 --- /dev/null +++ b/nl/nl_BE/nathalie/x_low/nl_BE-nathalie-x_low.onnx.json @@ -0,0 +1,420 @@ +{ + "audio": { + "sample_rate": 16000, + "quality": "x_low" + }, + "espeak": { + "voice": "nl" + }, + "inference": { + "noise_scale": 0.667, + "length_scale": 1, + "noise_w": 0.8 + }, + "phoneme_map": {}, + "phoneme_id_map": { + "_": [ + 0 + ], + "^": [ + 1 + ], + "$": [ + 2 + ], + " ": [ + 3 + ], + "!": [ + 4 + ], + "'": [ + 5 + ], + "(": [ + 6 + ], + ")": [ + 7 + ], + ",": [ + 8 + ], + "-": [ + 9 + ], + ".": [ + 10 + ], + ":": [ + 11 + ], + ";": [ + 12 + ], + "?": [ + 13 + ], + "a": [ + 14 + ], + "b": [ + 15 + ], + "c": [ + 16 + ], + "d": [ + 17 + ], + "e": [ + 18 + ], + "f": [ + 19 + ], + "h": [ + 20 + ], + "i": [ + 21 + ], + "j": [ + 22 + ], + "k": [ + 23 + ], + "l": [ + 24 + ], + "m": [ + 25 + ], + "n": [ + 26 + ], + "o": [ + 27 + ], + "p": [ + 28 + ], + "q": [ + 29 + ], + "r": [ + 30 + ], + "s": [ + 31 + ], + "t": [ + 32 + ], + "u": [ + 33 + ], + "v": [ + 34 + ], + "w": [ + 35 + ], + "x": [ + 36 + ], + "y": [ + 37 + ], + "z": [ + 38 + ], + "æ": [ + 39 + ], + "ç": [ + 40 + ], + "ð": [ + 41 + ], + "ø": [ + 42 + ], + "ħ": [ + 43 + ], + "ŋ": [ + 44 + ], + "œ": [ + 45 + ], + "ǀ": [ + 46 + ], + "ǁ": [ + 47 + ], + "ǂ": [ + 48 + ], + "ǃ": [ + 49 + ], + "ɐ": [ + 50 + ], + "ɑ": [ + 51 + ], + "ɒ": [ + 52 + ], + "ɓ": [ + 53 + ], + "ɔ": [ + 54 + ], + "ɕ": [ + 55 + ], + "ɖ": [ + 56 + ], + "ɗ": [ + 57 + ], + "ɘ": [ + 58 + ], + "ə": [ + 59 + ], + "ɚ": [ + 60 + ], + "ɛ": [ + 61 + ], + "ɜ": [ + 62 + ], + "ɞ": [ + 63 + ], + "ɟ": [ + 64 + ], + "ɠ": [ + 65 + ], + "ɡ": [ + 66 + ], + "ɢ": [ + 67 + ], + "ɣ": [ + 68 + ], + "ɤ": [ + 69 + ], + "ɥ": [ + 70 + ], + "ɦ": [ + 71 + ], + "ɧ": [ + 72 + ], + "ɨ": [ + 73 + ], + "ɪ": [ + 74 + ], + "ɫ": [ + 75 + ], + "ɬ": [ + 76 + ], + "ɭ": [ + 77 + ], + "ɮ": [ + 78 + ], + "ɯ": [ + 79 + ], + "ɰ": [ + 80 + ], + "ɱ": [ + 81 + ], + "ɲ": [ + 82 + ], + "ɳ": [ + 83 + ], + "ɴ": [ + 84 + ], + "ɵ": [ + 85 + ], + "ɶ": [ + 86 + ], + "ɸ": [ + 87 + ], + "ɹ": [ + 88 + ], + "ɺ": [ + 89 + ], + "ɻ": [ + 90 + ], + "ɽ": [ + 91 + ], + "ɾ": [ + 92 + ], + "ʀ": [ + 93 + ], + "ʁ": [ + 94 + ], + "ʂ": [ + 95 + ], + "ʃ": [ + 96 + ], + "ʄ": [ + 97 + ], + "ʈ": [ + 98 + ], + "ʉ": [ + 99 + ], + "ʊ": [ + 100 + ], + "ʋ": [ + 101 + ], + "ʌ": [ + 102 + ], + "ʍ": [ + 103 + ], + "ʎ": [ + 104 + ], + "ʏ": [ + 105 + ], + "ʐ": [ + 106 + ], + "ʑ": [ + 107 + ], + "ʒ": [ + 108 + ], + "ʔ": [ + 109 + ], + "ʕ": [ + 110 + ], + "ʘ": [ + 111 + ], + "ʙ": [ + 112 + ], + "ʛ": [ + 113 + ], + "ʜ": [ + 114 + ], + "ʝ": [ + 115 + ], + "ʟ": [ + 116 + ], + "ʡ": [ + 117 + ], + "ʢ": [ + 118 + ], + "ʲ": [ + 119 + ], + "ˈ": [ + 120 + ], + "ˌ": [ + 121 + ], + "ː": [ + 122 + ], + "ˑ": [ + 123 + ], + "˞": [ + 124 + ], + "β": [ + 125 + ], + "θ": [ + 126 + ], + "χ": [ + 127 + ], + "ᵻ": [ + 128 + ], + "ⱱ": [ + 129 + ] + }, + "num_symbols": 130, + "num_speakers": 1, + "speaker_id_map": {}, + "piper_version": "0.2.0", + "language": { + "code": "nl_BE", + "family": "nl", + "region": "BE", + "name_native": "Nederlands", + "name_english": "Dutch", + "country_english": "Belgium" + }, + "dataset": "nathalie" +} \ No newline at end of file diff --git a/nl/nl_BE/nathalie/x_low/samples/speaker_0.mp3 b/nl/nl_BE/nathalie/x_low/samples/speaker_0.mp3 new file mode 100644 index 0000000000000000000000000000000000000000..ee5afbd765d3229da31ddfff1916629b7df9ca5c Binary files /dev/null and b/nl/nl_BE/nathalie/x_low/samples/speaker_0.mp3 differ diff --git a/nl/nl_BE/rdh/medium/ALIASES b/nl/nl_BE/rdh/medium/ALIASES new file mode 100644 index 0000000000000000000000000000000000000000..2a91b573fab55d5654e5a86c3100d5f5dc8ca7a6 --- /dev/null +++ b/nl/nl_BE/rdh/medium/ALIASES @@ -0,0 +1 @@ +nl-rdh-medium diff --git a/nl/nl_BE/rdh/medium/MODEL_CARD b/nl/nl_BE/rdh/medium/MODEL_CARD new file mode 100644 index 0000000000000000000000000000000000000000..2d2c9715d0a9296117466adcebb145cbd4daefd4 --- /dev/null +++ b/nl/nl_BE/rdh/medium/MODEL_CARD @@ -0,0 +1,15 @@ +# Model card for rdh (medium) + +* Language: nl_BE (Dutch, Belgium) +* Speakers: 1 +* Quality: medium +* Samplerate: 22,050Hz + +## Dataset + +* URL: https://github.com/r-dh/dutch-vl-tts +* License: CC0 1.0 Universal + +## Training + +Trained from scratch. diff --git a/nl/nl_BE/rdh/medium/nl_BE-rdh-medium.onnx b/nl/nl_BE/rdh/medium/nl_BE-rdh-medium.onnx new file mode 100644 index 0000000000000000000000000000000000000000..5f0e2865b9f0bd01815aa60b34a4ab29485419fc --- /dev/null +++ b/nl/nl_BE/rdh/medium/nl_BE-rdh-medium.onnx @@ -0,0 +1,3 @@ +version https://git-lfs.github.com/spec/v1 +oid sha256:fb1a9d89ff53dbd510c5d10bd1365c36a5f76249c3ca603076646fd778003db1 +size 63104526 diff --git a/nl/nl_BE/rdh/medium/nl_BE-rdh-medium.onnx.json b/nl/nl_BE/rdh/medium/nl_BE-rdh-medium.onnx.json new file mode 100644 index 0000000000000000000000000000000000000000..0b63a34ad9246064ef6922b7522b80b38a385c3b --- /dev/null +++ b/nl/nl_BE/rdh/medium/nl_BE-rdh-medium.onnx.json @@ -0,0 +1,420 @@ +{ + "audio": { + "sample_rate": 22050, + "quality": "medium" + }, + "espeak": { + "voice": "nl" + }, + "inference": { + "noise_scale": 0.667, + "length_scale": 1, + "noise_w": 0.8 + }, + "phoneme_map": {}, + "phoneme_id_map": { + "_": [ + 0 + ], + "^": [ + 1 + ], + "$": [ + 2 + ], + " ": [ + 3 + ], + "!": [ + 4 + ], + "'": [ + 5 + ], + "(": [ + 6 + ], + ")": [ + 7 + ], + ",": [ + 8 + ], + "-": [ + 9 + ], + ".": [ + 10 + ], + ":": [ + 11 + ], + ";": [ + 12 + ], + "?": [ + 13 + ], + "a": [ + 14 + ], + "b": [ + 15 + ], + "c": [ + 16 + ], + "d": [ + 17 + ], + "e": [ + 18 + ], + "f": [ + 19 + ], + "h": [ + 20 + ], + "i": [ + 21 + ], + "j": [ + 22 + ], + "k": [ + 23 + ], + "l": [ + 24 + ], + "m": [ + 25 + ], + "n": [ + 26 + ], + "o": [ + 27 + ], + "p": [ + 28 + ], + "q": [ + 29 + ], + "r": [ + 30 + ], + "s": [ + 31 + ], + "t": [ + 32 + ], + "u": [ + 33 + ], + "v": [ + 34 + ], + "w": [ + 35 + ], + "x": [ + 36 + ], + "y": [ + 37 + ], + "z": [ + 38 + ], + "æ": [ + 39 + ], + "ç": [ + 40 + ], + "ð": [ + 41 + ], + "ø": [ + 42 + ], + "ħ": [ + 43 + ], + "ŋ": [ + 44 + ], + "œ": [ + 45 + ], + "ǀ": [ + 46 + ], + "ǁ": [ + 47 + ], + "ǂ": [ + 48 + ], + "ǃ": [ + 49 + ], + "ɐ": [ + 50 + ], + "ɑ": [ + 51 + ], + "ɒ": [ + 52 + ], + "ɓ": [ + 53 + ], + "ɔ": [ + 54 + ], + "ɕ": [ + 55 + ], + "ɖ": [ + 56 + ], + "ɗ": [ + 57 + ], + "ɘ": [ + 58 + ], + "ə": [ + 59 + ], + "ɚ": [ + 60 + ], + "ɛ": [ + 61 + ], + "ɜ": [ + 62 + ], + "ɞ": [ + 63 + ], + "ɟ": [ + 64 + ], + "ɠ": [ + 65 + ], + "ɡ": [ + 66 + ], + "ɢ": [ + 67 + ], + "ɣ": [ + 68 + ], + "ɤ": [ + 69 + ], + "ɥ": [ + 70 + ], + "ɦ": [ + 71 + ], + "ɧ": [ + 72 + ], + "ɨ": [ + 73 + ], + "ɪ": [ + 74 + ], + "ɫ": [ + 75 + ], + "ɬ": [ + 76 + ], + "ɭ": [ + 77 + ], + "ɮ": [ + 78 + ], + "ɯ": [ + 79 + ], + "ɰ": [ + 80 + ], + "ɱ": [ + 81 + ], + "ɲ": [ + 82 + ], + "ɳ": [ + 83 + ], + "ɴ": [ + 84 + ], + "ɵ": [ + 85 + ], + "ɶ": [ + 86 + ], + "ɸ": [ + 87 + ], + "ɹ": [ + 88 + ], + "ɺ": [ + 89 + ], + "ɻ": [ + 90 + ], + "ɽ": [ + 91 + ], + "ɾ": [ + 92 + ], + "ʀ": [ + 93 + ], + "ʁ": [ + 94 + ], + "ʂ": [ + 95 + ], + "ʃ": [ + 96 + ], + "ʄ": [ + 97 + ], + "ʈ": [ + 98 + ], + "ʉ": [ + 99 + ], + "ʊ": [ + 100 + ], + "ʋ": [ + 101 + ], + "ʌ": [ + 102 + ], + "ʍ": [ + 103 + ], + "ʎ": [ + 104 + ], + "ʏ": [ + 105 + ], + "ʐ": [ + 106 + ], + "ʑ": [ + 107 + ], + "ʒ": [ + 108 + ], + "ʔ": [ + 109 + ], + "ʕ": [ + 110 + ], + "ʘ": [ + 111 + ], + "ʙ": [ + 112 + ], + "ʛ": [ + 113 + ], + "ʜ": [ + 114 + ], + "ʝ": [ + 115 + ], + "ʟ": [ + 116 + ], + "ʡ": [ + 117 + ], + "ʢ": [ + 118 + ], + "ʲ": [ + 119 + ], + "ˈ": [ + 120 + ], + "ˌ": [ + 121 + ], + "ː": [ + 122 + ], + "ˑ": [ + 123 + ], + "˞": [ + 124 + ], + "β": [ + 125 + ], + "θ": [ + 126 + ], + "χ": [ + 127 + ], + "ᵻ": [ + 128 + ], + "ⱱ": [ + 129 + ] + }, + "num_symbols": 130, + "num_speakers": 1, + "speaker_id_map": {}, + "piper_version": "0.2.0", + "language": { + "code": "nl_BE", + "family": "nl", + "region": "BE", + "name_native": "Nederlands", + "name_english": "Dutch", + "country_english": "Belgium" + }, + "dataset": "rdh" +} \ No newline at end of file diff --git a/nl/nl_BE/rdh/medium/samples/speaker_0.mp3 b/nl/nl_BE/rdh/medium/samples/speaker_0.mp3 new file mode 100644 index 0000000000000000000000000000000000000000..22a795ba42f80b937ee57758bd4e689cb0b4a045 Binary files /dev/null and b/nl/nl_BE/rdh/medium/samples/speaker_0.mp3 differ diff --git a/nl/nl_BE/rdh/x_low/ALIASES b/nl/nl_BE/rdh/x_low/ALIASES new file mode 100644 index 0000000000000000000000000000000000000000..6c8be74ad015a7701fdd21312d7de38d536a17c8 --- /dev/null +++ b/nl/nl_BE/rdh/x_low/ALIASES @@ -0,0 +1 @@ +nl-rdh-x-low diff --git a/nl/nl_BE/rdh/x_low/MODEL_CARD b/nl/nl_BE/rdh/x_low/MODEL_CARD new file mode 100644 index 0000000000000000000000000000000000000000..4c3e8589d41d1c89a9c815906cd5147198e720cc --- /dev/null +++ b/nl/nl_BE/rdh/x_low/MODEL_CARD @@ -0,0 +1,15 @@ +# Model card for rdh (x_low) + +* Language: nl_BE (Dutch, Belgium) +* Speakers: 1 +* Quality: x_low +* Samplerate: 16,000Hz + +## Dataset + +* URL: https://github.com/r-dh/dutch-vl-tts +* License: CC0 1.0 Universal + +## Training + +Trained from scratch. diff --git a/nl/nl_BE/rdh/x_low/nl_BE-rdh-x_low.onnx b/nl/nl_BE/rdh/x_low/nl_BE-rdh-x_low.onnx new file mode 100644 index 0000000000000000000000000000000000000000..9cfdbc359942544db0a11768c1a57afd07b28342 --- /dev/null +++ b/nl/nl_BE/rdh/x_low/nl_BE-rdh-x_low.onnx @@ -0,0 +1,3 @@ +version https://git-lfs.github.com/spec/v1 +oid sha256:fe43651248219b4e816e47d9708f2812f2a09af99470ed822113f3e913430d52 +size 20628813 diff --git a/nl/nl_BE/rdh/x_low/nl_BE-rdh-x_low.onnx.json b/nl/nl_BE/rdh/x_low/nl_BE-rdh-x_low.onnx.json new file mode 100644 index 0000000000000000000000000000000000000000..161bea069ef8fa7a86dff5c14acd3c0b4267b590 --- /dev/null +++ b/nl/nl_BE/rdh/x_low/nl_BE-rdh-x_low.onnx.json @@ -0,0 +1,420 @@ +{ + "audio": { + "sample_rate": 16000, + "quality": "x_low" + }, + "espeak": { + "voice": "nl" + }, + "inference": { + "noise_scale": 0.667, + "length_scale": 1, + "noise_w": 0.8 + }, + "phoneme_map": {}, + "phoneme_id_map": { + "_": [ + 0 + ], + "^": [ + 1 + ], + "$": [ + 2 + ], + " ": [ + 3 + ], + "!": [ + 4 + ], + "'": [ + 5 + ], + "(": [ + 6 + ], + ")": [ + 7 + ], + ",": [ + 8 + ], + "-": [ + 9 + ], + ".": [ + 10 + ], + ":": [ + 11 + ], + ";": [ + 12 + ], + "?": [ + 13 + ], + "a": [ + 14 + ], + "b": [ + 15 + ], + "c": [ + 16 + ], + "d": [ + 17 + ], + "e": [ + 18 + ], + "f": [ + 19 + ], + "h": [ + 20 + ], + "i": [ + 21 + ], + "j": [ + 22 + ], + "k": [ + 23 + ], + "l": [ + 24 + ], + "m": [ + 25 + ], + "n": [ + 26 + ], + "o": [ + 27 + ], + "p": [ + 28 + ], + "q": [ + 29 + ], + "r": [ + 30 + ], + "s": [ + 31 + ], + "t": [ + 32 + ], + "u": [ + 33 + ], + "v": [ + 34 + ], + "w": [ + 35 + ], + "x": [ + 36 + ], + "y": [ + 37 + ], + "z": [ + 38 + ], + "æ": [ + 39 + ], + "ç": [ + 40 + ], + "ð": [ + 41 + ], + "ø": [ + 42 + ], + "ħ": [ + 43 + ], + "ŋ": [ + 44 + ], + "œ": [ + 45 + ], + "ǀ": [ + 46 + ], + "ǁ": [ + 47 + ], + "ǂ": [ + 48 + ], + "ǃ": [ + 49 + ], + "ɐ": [ + 50 + ], + "ɑ": [ + 51 + ], + "ɒ": [ + 52 + ], + "ɓ": [ + 53 + ], + "ɔ": [ + 54 + ], + "ɕ": [ + 55 + ], + "ɖ": [ + 56 + ], + "ɗ": [ + 57 + ], + "ɘ": [ + 58 + ], + "ə": [ + 59 + ], + "ɚ": [ + 60 + ], + "ɛ": [ + 61 + ], + "ɜ": [ + 62 + ], + "ɞ": [ + 63 + ], + "ɟ": [ + 64 + ], + "ɠ": [ + 65 + ], + "ɡ": [ + 66 + ], + "ɢ": [ + 67 + ], + "ɣ": [ + 68 + ], + "ɤ": [ + 69 + ], + "ɥ": [ + 70 + ], + "ɦ": [ + 71 + ], + "ɧ": [ + 72 + ], + "ɨ": [ + 73 + ], + "ɪ": [ + 74 + ], + "ɫ": [ + 75 + ], + "ɬ": [ + 76 + ], + "ɭ": [ + 77 + ], + "ɮ": [ + 78 + ], + "ɯ": [ + 79 + ], + "ɰ": [ + 80 + ], + "ɱ": [ + 81 + ], + "ɲ": [ + 82 + ], + "ɳ": [ + 83 + ], + "ɴ": [ + 84 + ], + "ɵ": [ + 85 + ], + "ɶ": [ + 86 + ], + "ɸ": [ + 87 + ], + "ɹ": [ + 88 + ], + "ɺ": [ + 89 + ], + "ɻ": [ + 90 + ], + "ɽ": [ + 91 + ], + "ɾ": [ + 92 + ], + "ʀ": [ + 93 + ], + "ʁ": [ + 94 + ], + "ʂ": [ + 95 + ], + "ʃ": [ + 96 + ], + "ʄ": [ + 97 + ], + "ʈ": [ + 98 + ], + "ʉ": [ + 99 + ], + "ʊ": [ + 100 + ], + "ʋ": [ + 101 + ], + "ʌ": [ + 102 + ], + "ʍ": [ + 103 + ], + "ʎ": [ + 104 + ], + "ʏ": [ + 105 + ], + "ʐ": [ + 106 + ], + "ʑ": [ + 107 + ], + "ʒ": [ + 108 + ], + "ʔ": [ + 109 + ], + "ʕ": [ + 110 + ], + "ʘ": [ + 111 + ], + "ʙ": [ + 112 + ], + "ʛ": [ + 113 + ], + "ʜ": [ + 114 + ], + "ʝ": [ + 115 + ], + "ʟ": [ + 116 + ], + "ʡ": [ + 117 + ], + "ʢ": [ + 118 + ], + "ʲ": [ + 119 + ], + "ˈ": [ + 120 + ], + "ˌ": [ + 121 + ], + "ː": [ + 122 + ], + "ˑ": [ + 123 + ], + "˞": [ + 124 + ], + "β": [ + 125 + ], + "θ": [ + 126 + ], + "χ": [ + 127 + ], + "ᵻ": [ + 128 + ], + "ⱱ": [ + 129 + ] + }, + "num_symbols": 130, + "num_speakers": 1, + "speaker_id_map": {}, + "piper_version": "0.2.0", + "language": { + "code": "nl_BE", + "family": "nl", + "region": "BE", + "name_native": "Nederlands", + "name_english": "Dutch", + "country_english": "Belgium" + }, + "dataset": "rdh" +} \ No newline at end of file diff --git a/nl/nl_BE/rdh/x_low/samples/speaker_0.mp3 b/nl/nl_BE/rdh/x_low/samples/speaker_0.mp3 new file mode 100644 index 0000000000000000000000000000000000000000..ffa3b07780c58857020f3de85b1527039cad22e6 Binary files /dev/null and b/nl/nl_BE/rdh/x_low/samples/speaker_0.mp3 differ diff --git a/nl/nl_NL/mls_5809/low/ALIASES b/nl/nl_NL/mls_5809/low/ALIASES new file mode 100644 index 0000000000000000000000000000000000000000..d582c7f5a2126f5e3c922921719ab30565387a75 --- /dev/null +++ b/nl/nl_NL/mls_5809/low/ALIASES @@ -0,0 +1 @@ +nl-mls_5809-low diff --git a/nl/nl_NL/mls_5809/low/MODEL_CARD b/nl/nl_NL/mls_5809/low/MODEL_CARD new file mode 100644 index 0000000000000000000000000000000000000000..3679777c385815b5067eb73c1a05725ca574ffb2 --- /dev/null +++ b/nl/nl_NL/mls_5809/low/MODEL_CARD @@ -0,0 +1,15 @@ +# Model card for mls_5809 (low) + +* Language: nl_NL (Dutch, Netherlands) +* Speakers: 1 +* Quality: low +* Samplerate: 16,000Hz + +## Dataset + +* URL: http://www.openslr.org/94/ +* License: CC BY 4.0 + +## Training + +Finetuned from U.S. English Ryan voice (low quality). diff --git a/nl/nl_NL/mls_5809/low/nl_NL-mls_5809-low.onnx b/nl/nl_NL/mls_5809/low/nl_NL-mls_5809-low.onnx new file mode 100644 index 0000000000000000000000000000000000000000..27edab369bc67560660d487ed7577999363b13df --- /dev/null +++ b/nl/nl_NL/mls_5809/low/nl_NL-mls_5809-low.onnx @@ -0,0 +1,3 @@ +version https://git-lfs.github.com/spec/v1 +oid sha256:b7860ada7fa27e5e29b08e7531b0e0d74b3f0602e466a22ff95f59f80a0c4327 +size 63104526 diff --git a/nl/nl_NL/mls_5809/low/nl_NL-mls_5809-low.onnx.json b/nl/nl_NL/mls_5809/low/nl_NL-mls_5809-low.onnx.json new file mode 100644 index 0000000000000000000000000000000000000000..25bbe393ca436e380890d2fb12c926d4eefbf84b --- /dev/null +++ b/nl/nl_NL/mls_5809/low/nl_NL-mls_5809-low.onnx.json @@ -0,0 +1,420 @@ +{ + "audio": { + "sample_rate": 16000, + "quality": "low" + }, + "espeak": { + "voice": "nl" + }, + "inference": { + "noise_scale": 0.667, + "length_scale": 1, + "noise_w": 0.8 + }, + "phoneme_map": {}, + "phoneme_id_map": { + "_": [ + 0 + ], + "^": [ + 1 + ], + "$": [ + 2 + ], + " ": [ + 3 + ], + "!": [ + 4 + ], + "'": [ + 5 + ], + "(": [ + 6 + ], + ")": [ + 7 + ], + ",": [ + 8 + ], + "-": [ + 9 + ], + ".": [ + 10 + ], + ":": [ + 11 + ], + ";": [ + 12 + ], + "?": [ + 13 + ], + "a": [ + 14 + ], + "b": [ + 15 + ], + "c": [ + 16 + ], + "d": [ + 17 + ], + "e": [ + 18 + ], + "f": [ + 19 + ], + "h": [ + 20 + ], + "i": [ + 21 + ], + "j": [ + 22 + ], + "k": [ + 23 + ], + "l": [ + 24 + ], + "m": [ + 25 + ], + "n": [ + 26 + ], + "o": [ + 27 + ], + "p": [ + 28 + ], + "q": [ + 29 + ], + "r": [ + 30 + ], + "s": [ + 31 + ], + "t": [ + 32 + ], + "u": [ + 33 + ], + "v": [ + 34 + ], + "w": [ + 35 + ], + "x": [ + 36 + ], + "y": [ + 37 + ], + "z": [ + 38 + ], + "æ": [ + 39 + ], + "ç": [ + 40 + ], + "ð": [ + 41 + ], + "ø": [ + 42 + ], + "ħ": [ + 43 + ], + "ŋ": [ + 44 + ], + "œ": [ + 45 + ], + "ǀ": [ + 46 + ], + "ǁ": [ + 47 + ], + "ǂ": [ + 48 + ], + "ǃ": [ + 49 + ], + "ɐ": [ + 50 + ], + "ɑ": [ + 51 + ], + "ɒ": [ + 52 + ], + "ɓ": [ + 53 + ], + "ɔ": [ + 54 + ], + "ɕ": [ + 55 + ], + "ɖ": [ + 56 + ], + "ɗ": [ + 57 + ], + "ɘ": [ + 58 + ], + "ə": [ + 59 + ], + "ɚ": [ + 60 + ], + "ɛ": [ + 61 + ], + "ɜ": [ + 62 + ], + "ɞ": [ + 63 + ], + "ɟ": [ + 64 + ], + "ɠ": [ + 65 + ], + "ɡ": [ + 66 + ], + "ɢ": [ + 67 + ], + "ɣ": [ + 68 + ], + "ɤ": [ + 69 + ], + "ɥ": [ + 70 + ], + "ɦ": [ + 71 + ], + "ɧ": [ + 72 + ], + "ɨ": [ + 73 + ], + "ɪ": [ + 74 + ], + "ɫ": [ + 75 + ], + "ɬ": [ + 76 + ], + "ɭ": [ + 77 + ], + "ɮ": [ + 78 + ], + "ɯ": [ + 79 + ], + "ɰ": [ + 80 + ], + "ɱ": [ + 81 + ], + "ɲ": [ + 82 + ], + "ɳ": [ + 83 + ], + "ɴ": [ + 84 + ], + "ɵ": [ + 85 + ], + "ɶ": [ + 86 + ], + "ɸ": [ + 87 + ], + "ɹ": [ + 88 + ], + "ɺ": [ + 89 + ], + "ɻ": [ + 90 + ], + "ɽ": [ + 91 + ], + "ɾ": [ + 92 + ], + "ʀ": [ + 93 + ], + "ʁ": [ + 94 + ], + "ʂ": [ + 95 + ], + "ʃ": [ + 96 + ], + "ʄ": [ + 97 + ], + "ʈ": [ + 98 + ], + "ʉ": [ + 99 + ], + "ʊ": [ + 100 + ], + "ʋ": [ + 101 + ], + "ʌ": [ + 102 + ], + "ʍ": [ + 103 + ], + "ʎ": [ + 104 + ], + "ʏ": [ + 105 + ], + "ʐ": [ + 106 + ], + "ʑ": [ + 107 + ], + "ʒ": [ + 108 + ], + "ʔ": [ + 109 + ], + "ʕ": [ + 110 + ], + "ʘ": [ + 111 + ], + "ʙ": [ + 112 + ], + "ʛ": [ + 113 + ], + "ʜ": [ + 114 + ], + "ʝ": [ + 115 + ], + "ʟ": [ + 116 + ], + "ʡ": [ + 117 + ], + "ʢ": [ + 118 + ], + "ʲ": [ + 119 + ], + "ˈ": [ + 120 + ], + "ˌ": [ + 121 + ], + "ː": [ + 122 + ], + "ˑ": [ + 123 + ], + "˞": [ + 124 + ], + "β": [ + 125 + ], + "θ": [ + 126 + ], + "χ": [ + 127 + ], + "ᵻ": [ + 128 + ], + "ⱱ": [ + 129 + ] + }, + "num_symbols": 130, + "num_speakers": 1, + "speaker_id_map": {}, + "piper_version": "0.2.0", + "language": { + "code": "nl_NL", + "family": "nl", + "region": "NL", + "name_native": "Nederlands", + "name_english": "Dutch", + "country_english": "Netherlands" + }, + "dataset": "mls_5809" +} \ No newline at end of file diff --git a/nl/nl_NL/mls_5809/low/samples/speaker_0.mp3 b/nl/nl_NL/mls_5809/low/samples/speaker_0.mp3 new file mode 100644 index 0000000000000000000000000000000000000000..d2b1b895db41d29ea783abb8b3b1870404f68972 Binary files /dev/null and b/nl/nl_NL/mls_5809/low/samples/speaker_0.mp3 differ diff --git a/nl/nl_NL/mls_7432/low/ALIASES b/nl/nl_NL/mls_7432/low/ALIASES new file mode 100644 index 0000000000000000000000000000000000000000..69b52ab2e6a963bb629c6dbe8922cdd3dfe087c1 --- /dev/null +++ b/nl/nl_NL/mls_7432/low/ALIASES @@ -0,0 +1 @@ +nl-mls_7432-low diff --git a/nl/nl_NL/mls_7432/low/MODEL_CARD b/nl/nl_NL/mls_7432/low/MODEL_CARD new file mode 100644 index 0000000000000000000000000000000000000000..c34092f687247d7bf5d7d9fd77d12794e80fb4a5 --- /dev/null +++ b/nl/nl_NL/mls_7432/low/MODEL_CARD @@ -0,0 +1,15 @@ +# Model card for mls_7432 (low) + +* Language: nl_NL (Dutch, Netherlands) +* Speakers: 1 +* Quality: low +* Samplerate: 16,000Hz + +## Dataset + +* URL: http://www.openslr.org/94/ +* License: CC BY 4.0 + +## Training + +Finetuned from U.S. English Ryan voice (low quality). diff --git a/nl/nl_NL/mls_7432/low/nl_NL-mls_7432-low.onnx b/nl/nl_NL/mls_7432/low/nl_NL-mls_7432-low.onnx new file mode 100644 index 0000000000000000000000000000000000000000..678725f113ecd48ea117a9aa5e800ca7c95097fa --- /dev/null +++ b/nl/nl_NL/mls_7432/low/nl_NL-mls_7432-low.onnx @@ -0,0 +1,3 @@ +version https://git-lfs.github.com/spec/v1 +oid sha256:02feb278732c77f34c4e106d0cdc446e7e577ce29c5d7051da85a9c241ee42f0 +size 63104526 diff --git a/nl/nl_NL/mls_7432/low/nl_NL-mls_7432-low.onnx.json b/nl/nl_NL/mls_7432/low/nl_NL-mls_7432-low.onnx.json new file mode 100644 index 0000000000000000000000000000000000000000..b1e673bb354da4293e089daeee97adbe4ec93ae6 --- /dev/null +++ b/nl/nl_NL/mls_7432/low/nl_NL-mls_7432-low.onnx.json @@ -0,0 +1,420 @@ +{ + "audio": { + "sample_rate": 16000, + "quality": "low" + }, + "espeak": { + "voice": "nl" + }, + "inference": { + "noise_scale": 0.667, + "length_scale": 1, + "noise_w": 0.8 + }, + "phoneme_map": {}, + "phoneme_id_map": { + "_": [ + 0 + ], + "^": [ + 1 + ], + "$": [ + 2 + ], + " ": [ + 3 + ], + "!": [ + 4 + ], + "'": [ + 5 + ], + "(": [ + 6 + ], + ")": [ + 7 + ], + ",": [ + 8 + ], + "-": [ + 9 + ], + ".": [ + 10 + ], + ":": [ + 11 + ], + ";": [ + 12 + ], + "?": [ + 13 + ], + "a": [ + 14 + ], + "b": [ + 15 + ], + "c": [ + 16 + ], + "d": [ + 17 + ], + "e": [ + 18 + ], + "f": [ + 19 + ], + "h": [ + 20 + ], + "i": [ + 21 + ], + "j": [ + 22 + ], + "k": [ + 23 + ], + "l": [ + 24 + ], + "m": [ + 25 + ], + "n": [ + 26 + ], + "o": [ + 27 + ], + "p": [ + 28 + ], + "q": [ + 29 + ], + "r": [ + 30 + ], + "s": [ + 31 + ], + "t": [ + 32 + ], + "u": [ + 33 + ], + "v": [ + 34 + ], + "w": [ + 35 + ], + "x": [ + 36 + ], + "y": [ + 37 + ], + "z": [ + 38 + ], + "æ": [ + 39 + ], + "ç": [ + 40 + ], + "ð": [ + 41 + ], + "ø": [ + 42 + ], + "ħ": [ + 43 + ], + "ŋ": [ + 44 + ], + "œ": [ + 45 + ], + "ǀ": [ + 46 + ], + "ǁ": [ + 47 + ], + "ǂ": [ + 48 + ], + "ǃ": [ + 49 + ], + "ɐ": [ + 50 + ], + "ɑ": [ + 51 + ], + "ɒ": [ + 52 + ], + "ɓ": [ + 53 + ], + "ɔ": [ + 54 + ], + "ɕ": [ + 55 + ], + "ɖ": [ + 56 + ], + "ɗ": [ + 57 + ], + "ɘ": [ + 58 + ], + "ə": [ + 59 + ], + "ɚ": [ + 60 + ], + "ɛ": [ + 61 + ], + "ɜ": [ + 62 + ], + "ɞ": [ + 63 + ], + "ɟ": [ + 64 + ], + "ɠ": [ + 65 + ], + "ɡ": [ + 66 + ], + "ɢ": [ + 67 + ], + "ɣ": [ + 68 + ], + "ɤ": [ + 69 + ], + "ɥ": [ + 70 + ], + "ɦ": [ + 71 + ], + "ɧ": [ + 72 + ], + "ɨ": [ + 73 + ], + "ɪ": [ + 74 + ], + "ɫ": [ + 75 + ], + "ɬ": [ + 76 + ], + "ɭ": [ + 77 + ], + "ɮ": [ + 78 + ], + "ɯ": [ + 79 + ], + "ɰ": [ + 80 + ], + "ɱ": [ + 81 + ], + "ɲ": [ + 82 + ], + "ɳ": [ + 83 + ], + "ɴ": [ + 84 + ], + "ɵ": [ + 85 + ], + "ɶ": [ + 86 + ], + "ɸ": [ + 87 + ], + "ɹ": [ + 88 + ], + "ɺ": [ + 89 + ], + "ɻ": [ + 90 + ], + "ɽ": [ + 91 + ], + "ɾ": [ + 92 + ], + "ʀ": [ + 93 + ], + "ʁ": [ + 94 + ], + "ʂ": [ + 95 + ], + "ʃ": [ + 96 + ], + "ʄ": [ + 97 + ], + "ʈ": [ + 98 + ], + "ʉ": [ + 99 + ], + "ʊ": [ + 100 + ], + "ʋ": [ + 101 + ], + "ʌ": [ + 102 + ], + "ʍ": [ + 103 + ], + "ʎ": [ + 104 + ], + "ʏ": [ + 105 + ], + "ʐ": [ + 106 + ], + "ʑ": [ + 107 + ], + "ʒ": [ + 108 + ], + "ʔ": [ + 109 + ], + "ʕ": [ + 110 + ], + "ʘ": [ + 111 + ], + "ʙ": [ + 112 + ], + "ʛ": [ + 113 + ], + "ʜ": [ + 114 + ], + "ʝ": [ + 115 + ], + "ʟ": [ + 116 + ], + "ʡ": [ + 117 + ], + "ʢ": [ + 118 + ], + "ʲ": [ + 119 + ], + "ˈ": [ + 120 + ], + "ˌ": [ + 121 + ], + "ː": [ + 122 + ], + "ˑ": [ + 123 + ], + "˞": [ + 124 + ], + "β": [ + 125 + ], + "θ": [ + 126 + ], + "χ": [ + 127 + ], + "ᵻ": [ + 128 + ], + "ⱱ": [ + 129 + ] + }, + "num_symbols": 130, + "num_speakers": 1, + "speaker_id_map": {}, + "piper_version": "0.2.0", + "language": { + "code": "nl_NL", + "family": "nl", + "region": "NL", + "name_native": "Nederlands", + "name_english": "Dutch", + "country_english": "Netherlands" + }, + "dataset": "mls_7432" +} \ No newline at end of file diff --git a/nl/nl_NL/mls_7432/low/samples/speaker_0.mp3 b/nl/nl_NL/mls_7432/low/samples/speaker_0.mp3 new file mode 100644 index 0000000000000000000000000000000000000000..f7f79ca344afba4406982ccd9143e609aaf62784 Binary files /dev/null and b/nl/nl_NL/mls_7432/low/samples/speaker_0.mp3 differ diff --git a/no/no_NO/talesyntese/medium/ALIASES b/no/no_NO/talesyntese/medium/ALIASES new file mode 100644 index 0000000000000000000000000000000000000000..cca708a3078f673bfed02a127a0f9e39d56d491f --- /dev/null +++ b/no/no_NO/talesyntese/medium/ALIASES @@ -0,0 +1 @@ +no-talesyntese-medium diff --git a/no/no_NO/talesyntese/medium/MODEL_CARD b/no/no_NO/talesyntese/medium/MODEL_CARD new file mode 100644 index 0000000000000000000000000000000000000000..31d48e4eaa4ef4988dba357fe1442552e0cd8fd7 --- /dev/null +++ b/no/no_NO/talesyntese/medium/MODEL_CARD @@ -0,0 +1,15 @@ +# Model card for talesyntese (medium) + +* Language: no_NO (Norwegian, Norway) +* Speakers: 1 +* Quality: medium +* Samplerate: 22,050Hz + +## Dataset + +* URL: https://www.nb.no/sprakbanken/en/resource-catalogue/oai-nb-no-sbr-15/ +* License: CC0 + +## Training + +Fine-tuned from U.S. English lessac voice (medium quality). diff --git a/no/no_NO/talesyntese/medium/no_NO-talesyntese-medium.onnx b/no/no_NO/talesyntese/medium/no_NO-talesyntese-medium.onnx new file mode 100644 index 0000000000000000000000000000000000000000..9b6bd3afe71a0bdc684ba8c1ae3255328174bc2e --- /dev/null +++ b/no/no_NO/talesyntese/medium/no_NO-talesyntese-medium.onnx @@ -0,0 +1,3 @@ +version https://git-lfs.github.com/spec/v1 +oid sha256:b763aebe02e72c4628c4074c4b8b448f05ffe523b975bfcd58e2c5d531270c13 +size 63201294 diff --git a/no/no_NO/talesyntese/medium/no_NO-talesyntese-medium.onnx.json b/no/no_NO/talesyntese/medium/no_NO-talesyntese-medium.onnx.json new file mode 100644 index 0000000000000000000000000000000000000000..fa4c6f86da3e284fab1c91d88f66c0dd42e587ec --- /dev/null +++ b/no/no_NO/talesyntese/medium/no_NO-talesyntese-medium.onnx.json @@ -0,0 +1,493 @@ +{ + "audio": { + "sample_rate": 22050, + "quality": "medium" + }, + "espeak": { + "voice": "nb" + }, + "inference": { + "noise_scale": 0.667, + "length_scale": 1, + "noise_w": 0.8 + }, + "phoneme_type": "espeak", + "phoneme_map": {}, + "phoneme_id_map": { + "_": [ + 0 + ], + "^": [ + 1 + ], + "$": [ + 2 + ], + " ": [ + 3 + ], + "!": [ + 4 + ], + "'": [ + 5 + ], + "(": [ + 6 + ], + ")": [ + 7 + ], + ",": [ + 8 + ], + "-": [ + 9 + ], + ".": [ + 10 + ], + ":": [ + 11 + ], + ";": [ + 12 + ], + "?": [ + 13 + ], + "a": [ + 14 + ], + "b": [ + 15 + ], + "c": [ + 16 + ], + "d": [ + 17 + ], + "e": [ + 18 + ], + "f": [ + 19 + ], + "h": [ + 20 + ], + "i": [ + 21 + ], + "j": [ + 22 + ], + "k": [ + 23 + ], + "l": [ + 24 + ], + "m": [ + 25 + ], + "n": [ + 26 + ], + "o": [ + 27 + ], + "p": [ + 28 + ], + "q": [ + 29 + ], + "r": [ + 30 + ], + "s": [ + 31 + ], + "t": [ + 32 + ], + "u": [ + 33 + ], + "v": [ + 34 + ], + "w": [ + 35 + ], + "x": [ + 36 + ], + "y": [ + 37 + ], + "z": [ + 38 + ], + "æ": [ + 39 + ], + "ç": [ + 40 + ], + "ð": [ + 41 + ], + "ø": [ + 42 + ], + "ħ": [ + 43 + ], + "ŋ": [ + 44 + ], + "œ": [ + 45 + ], + "ǀ": [ + 46 + ], + "ǁ": [ + 47 + ], + "ǂ": [ + 48 + ], + "ǃ": [ + 49 + ], + "ɐ": [ + 50 + ], + "ɑ": [ + 51 + ], + "ɒ": [ + 52 + ], + "ɓ": [ + 53 + ], + "ɔ": [ + 54 + ], + "ɕ": [ + 55 + ], + "ɖ": [ + 56 + ], + "ɗ": [ + 57 + ], + "ɘ": [ + 58 + ], + "ə": [ + 59 + ], + "ɚ": [ + 60 + ], + "ɛ": [ + 61 + ], + "ɜ": [ + 62 + ], + "ɞ": [ + 63 + ], + "ɟ": [ + 64 + ], + "ɠ": [ + 65 + ], + "ɡ": [ + 66 + ], + "ɢ": [ + 67 + ], + "ɣ": [ + 68 + ], + "ɤ": [ + 69 + ], + "ɥ": [ + 70 + ], + "ɦ": [ + 71 + ], + "ɧ": [ + 72 + ], + "ɨ": [ + 73 + ], + "ɪ": [ + 74 + ], + "ɫ": [ + 75 + ], + "ɬ": [ + 76 + ], + "ɭ": [ + 77 + ], + "ɮ": [ + 78 + ], + "ɯ": [ + 79 + ], + "ɰ": [ + 80 + ], + "ɱ": [ + 81 + ], + "ɲ": [ + 82 + ], + "ɳ": [ + 83 + ], + "ɴ": [ + 84 + ], + "ɵ": [ + 85 + ], + "ɶ": [ + 86 + ], + "ɸ": [ + 87 + ], + "ɹ": [ + 88 + ], + "ɺ": [ + 89 + ], + "ɻ": [ + 90 + ], + "ɽ": [ + 91 + ], + "ɾ": [ + 92 + ], + "ʀ": [ + 93 + ], + "ʁ": [ + 94 + ], + "ʂ": [ + 95 + ], + "ʃ": [ + 96 + ], + "ʄ": [ + 97 + ], + "ʈ": [ + 98 + ], + "ʉ": [ + 99 + ], + "ʊ": [ + 100 + ], + "ʋ": [ + 101 + ], + "ʌ": [ + 102 + ], + "ʍ": [ + 103 + ], + "ʎ": [ + 104 + ], + "ʏ": [ + 105 + ], + "ʐ": [ + 106 + ], + "ʑ": [ + 107 + ], + "ʒ": [ + 108 + ], + "ʔ": [ + 109 + ], + "ʕ": [ + 110 + ], + "ʘ": [ + 111 + ], + "ʙ": [ + 112 + ], + "ʛ": [ + 113 + ], + "ʜ": [ + 114 + ], + "ʝ": [ + 115 + ], + "ʟ": [ + 116 + ], + "ʡ": [ + 117 + ], + "ʢ": [ + 118 + ], + "ʲ": [ + 119 + ], + "ˈ": [ + 120 + ], + "ˌ": [ + 121 + ], + "ː": [ + 122 + ], + "ˑ": [ + 123 + ], + "˞": [ + 124 + ], + "β": [ + 125 + ], + "θ": [ + 126 + ], + "χ": [ + 127 + ], + "ᵻ": [ + 128 + ], + "ⱱ": [ + 129 + ], + "0": [ + 130 + ], + "1": [ + 131 + ], + "2": [ + 132 + ], + "3": [ + 133 + ], + "4": [ + 134 + ], + "5": [ + 135 + ], + "6": [ + 136 + ], + "7": [ + 137 + ], + "8": [ + 138 + ], + "9": [ + 139 + ], + "̧": [ + 140 + ], + "̃": [ + 141 + ], + "̪": [ + 142 + ], + "̯": [ + 143 + ], + "̩": [ + 144 + ], + "ʰ": [ + 145 + ], + "ˤ": [ + 146 + ], + "ε": [ + 147 + ], + "↓": [ + 148 + ], + "#": [ + 149 + ], + "\"": [ + 150 + ], + "↑": [ + 151 + ], + "̺": [ + 152 + ], + "̻": [ + 153 + ] + }, + "num_symbols": 256, + "num_speakers": 1, + "speaker_id_map": {}, + "piper_version": "1.0.0", + "language": { + "code": "no_NO", + "family": "no", + "region": "NO", + "name_native": "Norsk", + "name_english": "Norwegian", + "country_english": "Norway" + }, + "dataset": "talesyntese" +} \ No newline at end of file diff --git a/no/no_NO/talesyntese/medium/samples/speaker_0.mp3 b/no/no_NO/talesyntese/medium/samples/speaker_0.mp3 new file mode 100644 index 0000000000000000000000000000000000000000..45981f81d46fb0041ef65e72d04f8406c5f4793e Binary files /dev/null and b/no/no_NO/talesyntese/medium/samples/speaker_0.mp3 differ diff --git a/pl/pl_PL/darkman/medium/MODEL_CARD b/pl/pl_PL/darkman/medium/MODEL_CARD new file mode 100644 index 0000000000000000000000000000000000000000..36e4b5617fdd97e27ce334b04681ddbf6773c5ad --- /dev/null +++ b/pl/pl_PL/darkman/medium/MODEL_CARD @@ -0,0 +1,15 @@ +# Model card for darkman (medium) + +* Language: pl_PL (Polish, Poland) +* Speakers: 1 +* Quality: medium +* Samplerate: 22,050Hz + +## Dataset + +* URL: https://github.com/NabuCasa/voice-datasets +* License: CC0 + +## Training + +Finetuned from U.S. English lessac voice (medium quality). diff --git a/pl/pl_PL/darkman/medium/pl_PL-darkman-medium.onnx b/pl/pl_PL/darkman/medium/pl_PL-darkman-medium.onnx new file mode 100644 index 0000000000000000000000000000000000000000..a65f75aeb7b097a197ecc31151a89a38951679f4 --- /dev/null +++ b/pl/pl_PL/darkman/medium/pl_PL-darkman-medium.onnx @@ -0,0 +1,3 @@ +version https://git-lfs.github.com/spec/v1 +oid sha256:db505438a5364e8e2e0242c4324130a873ed660dfbe8d9689cef428ffb1b645f +size 63201294 diff --git a/pl/pl_PL/darkman/medium/pl_PL-darkman-medium.onnx.json b/pl/pl_PL/darkman/medium/pl_PL-darkman-medium.onnx.json new file mode 100644 index 0000000000000000000000000000000000000000..d7973acf52d6b18a90298f73d14178b047766bdd --- /dev/null +++ b/pl/pl_PL/darkman/medium/pl_PL-darkman-medium.onnx.json @@ -0,0 +1,487 @@ +{ + "audio": { + "sample_rate": 22050, + "quality": "medium" + }, + "espeak": { + "voice": "pl" + }, + "inference": { + "noise_scale": 0.667, + "length_scale": 1, + "noise_w": 0.8 + }, + "phoneme_type": "espeak", + "phoneme_map": {}, + "phoneme_id_map": { + "_": [ + 0 + ], + "^": [ + 1 + ], + "$": [ + 2 + ], + " ": [ + 3 + ], + "!": [ + 4 + ], + "'": [ + 5 + ], + "(": [ + 6 + ], + ")": [ + 7 + ], + ",": [ + 8 + ], + "-": [ + 9 + ], + ".": [ + 10 + ], + ":": [ + 11 + ], + ";": [ + 12 + ], + "?": [ + 13 + ], + "a": [ + 14 + ], + "b": [ + 15 + ], + "c": [ + 16 + ], + "d": [ + 17 + ], + "e": [ + 18 + ], + "f": [ + 19 + ], + "h": [ + 20 + ], + "i": [ + 21 + ], + "j": [ + 22 + ], + "k": [ + 23 + ], + "l": [ + 24 + ], + "m": [ + 25 + ], + "n": [ + 26 + ], + "o": [ + 27 + ], + "p": [ + 28 + ], + "q": [ + 29 + ], + "r": [ + 30 + ], + "s": [ + 31 + ], + "t": [ + 32 + ], + "u": [ + 33 + ], + "v": [ + 34 + ], + "w": [ + 35 + ], + "x": [ + 36 + ], + "y": [ + 37 + ], + "z": [ + 38 + ], + "æ": [ + 39 + ], + "ç": [ + 40 + ], + "ð": [ + 41 + ], + "ø": [ + 42 + ], + "ħ": [ + 43 + ], + "ŋ": [ + 44 + ], + "œ": [ + 45 + ], + "ǀ": [ + 46 + ], + "ǁ": [ + 47 + ], + "ǂ": [ + 48 + ], + "ǃ": [ + 49 + ], + "ɐ": [ + 50 + ], + "ɑ": [ + 51 + ], + "ɒ": [ + 52 + ], + "ɓ": [ + 53 + ], + "ɔ": [ + 54 + ], + "ɕ": [ + 55 + ], + "ɖ": [ + 56 + ], + "ɗ": [ + 57 + ], + "ɘ": [ + 58 + ], + "ə": [ + 59 + ], + "ɚ": [ + 60 + ], + "ɛ": [ + 61 + ], + "ɜ": [ + 62 + ], + "ɞ": [ + 63 + ], + "ɟ": [ + 64 + ], + "ɠ": [ + 65 + ], + "ɡ": [ + 66 + ], + "ɢ": [ + 67 + ], + "ɣ": [ + 68 + ], + "ɤ": [ + 69 + ], + "ɥ": [ + 70 + ], + "ɦ": [ + 71 + ], + "ɧ": [ + 72 + ], + "ɨ": [ + 73 + ], + "ɪ": [ + 74 + ], + "ɫ": [ + 75 + ], + "ɬ": [ + 76 + ], + "ɭ": [ + 77 + ], + "ɮ": [ + 78 + ], + "ɯ": [ + 79 + ], + "ɰ": [ + 80 + ], + "ɱ": [ + 81 + ], + "ɲ": [ + 82 + ], + "ɳ": [ + 83 + ], + "ɴ": [ + 84 + ], + "ɵ": [ + 85 + ], + "ɶ": [ + 86 + ], + "ɸ": [ + 87 + ], + "ɹ": [ + 88 + ], + "ɺ": [ + 89 + ], + "ɻ": [ + 90 + ], + "ɽ": [ + 91 + ], + "ɾ": [ + 92 + ], + "ʀ": [ + 93 + ], + "ʁ": [ + 94 + ], + "ʂ": [ + 95 + ], + "ʃ": [ + 96 + ], + "ʄ": [ + 97 + ], + "ʈ": [ + 98 + ], + "ʉ": [ + 99 + ], + "ʊ": [ + 100 + ], + "ʋ": [ + 101 + ], + "ʌ": [ + 102 + ], + "ʍ": [ + 103 + ], + "ʎ": [ + 104 + ], + "ʏ": [ + 105 + ], + "ʐ": [ + 106 + ], + "ʑ": [ + 107 + ], + "ʒ": [ + 108 + ], + "ʔ": [ + 109 + ], + "ʕ": [ + 110 + ], + "ʘ": [ + 111 + ], + "ʙ": [ + 112 + ], + "ʛ": [ + 113 + ], + "ʜ": [ + 114 + ], + "ʝ": [ + 115 + ], + "ʟ": [ + 116 + ], + "ʡ": [ + 117 + ], + "ʢ": [ + 118 + ], + "ʲ": [ + 119 + ], + "ˈ": [ + 120 + ], + "ˌ": [ + 121 + ], + "ː": [ + 122 + ], + "ˑ": [ + 123 + ], + "˞": [ + 124 + ], + "β": [ + 125 + ], + "θ": [ + 126 + ], + "χ": [ + 127 + ], + "ᵻ": [ + 128 + ], + "ⱱ": [ + 129 + ], + "0": [ + 130 + ], + "1": [ + 131 + ], + "2": [ + 132 + ], + "3": [ + 133 + ], + "4": [ + 134 + ], + "5": [ + 135 + ], + "6": [ + 136 + ], + "7": [ + 137 + ], + "8": [ + 138 + ], + "9": [ + 139 + ], + "̧": [ + 140 + ], + "̃": [ + 141 + ], + "̪": [ + 142 + ], + "̯": [ + 143 + ], + "̩": [ + 144 + ], + "ʰ": [ + 145 + ], + "ˤ": [ + 146 + ], + "ε": [ + 147 + ], + "↓": [ + 148 + ], + "#": [ + 149 + ], + "\"": [ + 150 + ], + "↑": [ + 151 + ] + }, + "num_symbols": 256, + "num_speakers": 1, + "speaker_id_map": {}, + "piper_version": "1.0.0", + "language": { + "code": "pl_PL", + "family": "pl", + "region": "PL", + "name_native": "Polski", + "name_english": "Polish", + "country_english": "Poland" + }, + "dataset": "darkman" +} \ No newline at end of file diff --git a/pl/pl_PL/darkman/medium/samples/speaker_0.mp3 b/pl/pl_PL/darkman/medium/samples/speaker_0.mp3 new file mode 100644 index 0000000000000000000000000000000000000000..7e1d9b6d0461c41346c4cf18022af9891aced79d Binary files /dev/null and b/pl/pl_PL/darkman/medium/samples/speaker_0.mp3 differ diff --git a/pl/pl_PL/gosia/medium/MODEL_CARD b/pl/pl_PL/gosia/medium/MODEL_CARD new file mode 100644 index 0000000000000000000000000000000000000000..ece4edd11b1ff75e7d67f6258794fe99cf4be2f6 --- /dev/null +++ b/pl/pl_PL/gosia/medium/MODEL_CARD @@ -0,0 +1,15 @@ +# Model card for gosia (medium) + +* Language: pl_PL (Polish, Poland) +* Speakers: 1 +* Quality: medium +* Samplerate: 22,050Hz + +## Dataset + +* URL: https://github.com/NabuCasa/voice-datasets +* License: CC0 + +## Training + +Finetuned from U.S. English lessac voice (medium quality). diff --git a/pl/pl_PL/gosia/medium/pl_PL-gosia-medium.onnx b/pl/pl_PL/gosia/medium/pl_PL-gosia-medium.onnx new file mode 100644 index 0000000000000000000000000000000000000000..bebc4ab63fb2dfc1a87d7ff4f8536f4c10034b25 --- /dev/null +++ b/pl/pl_PL/gosia/medium/pl_PL-gosia-medium.onnx @@ -0,0 +1,3 @@ +version https://git-lfs.github.com/spec/v1 +oid sha256:38f66464240ed74f186e6b7dc13c6e3b22e023426299f25c2b3cc9dfa9373fbc +size 63201294 diff --git a/pl/pl_PL/gosia/medium/pl_PL-gosia-medium.onnx.json b/pl/pl_PL/gosia/medium/pl_PL-gosia-medium.onnx.json new file mode 100644 index 0000000000000000000000000000000000000000..b8feaf17f485dfa6146a88f1a5812baea8a62860 --- /dev/null +++ b/pl/pl_PL/gosia/medium/pl_PL-gosia-medium.onnx.json @@ -0,0 +1,487 @@ +{ + "audio": { + "sample_rate": 22050, + "quality": "medium" + }, + "espeak": { + "voice": "pl" + }, + "inference": { + "noise_scale": 0.667, + "length_scale": 1, + "noise_w": 0.8 + }, + "phoneme_type": "espeak", + "phoneme_map": {}, + "phoneme_id_map": { + "_": [ + 0 + ], + "^": [ + 1 + ], + "$": [ + 2 + ], + " ": [ + 3 + ], + "!": [ + 4 + ], + "'": [ + 5 + ], + "(": [ + 6 + ], + ")": [ + 7 + ], + ",": [ + 8 + ], + "-": [ + 9 + ], + ".": [ + 10 + ], + ":": [ + 11 + ], + ";": [ + 12 + ], + "?": [ + 13 + ], + "a": [ + 14 + ], + "b": [ + 15 + ], + "c": [ + 16 + ], + "d": [ + 17 + ], + "e": [ + 18 + ], + "f": [ + 19 + ], + "h": [ + 20 + ], + "i": [ + 21 + ], + "j": [ + 22 + ], + "k": [ + 23 + ], + "l": [ + 24 + ], + "m": [ + 25 + ], + "n": [ + 26 + ], + "o": [ + 27 + ], + "p": [ + 28 + ], + "q": [ + 29 + ], + "r": [ + 30 + ], + "s": [ + 31 + ], + "t": [ + 32 + ], + "u": [ + 33 + ], + "v": [ + 34 + ], + "w": [ + 35 + ], + "x": [ + 36 + ], + "y": [ + 37 + ], + "z": [ + 38 + ], + "æ": [ + 39 + ], + "ç": [ + 40 + ], + "ð": [ + 41 + ], + "ø": [ + 42 + ], + "ħ": [ + 43 + ], + "ŋ": [ + 44 + ], + "œ": [ + 45 + ], + "ǀ": [ + 46 + ], + "ǁ": [ + 47 + ], + "ǂ": [ + 48 + ], + "ǃ": [ + 49 + ], + "ɐ": [ + 50 + ], + "ɑ": [ + 51 + ], + "ɒ": [ + 52 + ], + "ɓ": [ + 53 + ], + "ɔ": [ + 54 + ], + "ɕ": [ + 55 + ], + "ɖ": [ + 56 + ], + "ɗ": [ + 57 + ], + "ɘ": [ + 58 + ], + "ə": [ + 59 + ], + "ɚ": [ + 60 + ], + "ɛ": [ + 61 + ], + "ɜ": [ + 62 + ], + "ɞ": [ + 63 + ], + "ɟ": [ + 64 + ], + "ɠ": [ + 65 + ], + "ɡ": [ + 66 + ], + "ɢ": [ + 67 + ], + "ɣ": [ + 68 + ], + "ɤ": [ + 69 + ], + "ɥ": [ + 70 + ], + "ɦ": [ + 71 + ], + "ɧ": [ + 72 + ], + "ɨ": [ + 73 + ], + "ɪ": [ + 74 + ], + "ɫ": [ + 75 + ], + "ɬ": [ + 76 + ], + "ɭ": [ + 77 + ], + "ɮ": [ + 78 + ], + "ɯ": [ + 79 + ], + "ɰ": [ + 80 + ], + "ɱ": [ + 81 + ], + "ɲ": [ + 82 + ], + "ɳ": [ + 83 + ], + "ɴ": [ + 84 + ], + "ɵ": [ + 85 + ], + "ɶ": [ + 86 + ], + "ɸ": [ + 87 + ], + "ɹ": [ + 88 + ], + "ɺ": [ + 89 + ], + "ɻ": [ + 90 + ], + "ɽ": [ + 91 + ], + "ɾ": [ + 92 + ], + "ʀ": [ + 93 + ], + "ʁ": [ + 94 + ], + "ʂ": [ + 95 + ], + "ʃ": [ + 96 + ], + "ʄ": [ + 97 + ], + "ʈ": [ + 98 + ], + "ʉ": [ + 99 + ], + "ʊ": [ + 100 + ], + "ʋ": [ + 101 + ], + "ʌ": [ + 102 + ], + "ʍ": [ + 103 + ], + "ʎ": [ + 104 + ], + "ʏ": [ + 105 + ], + "ʐ": [ + 106 + ], + "ʑ": [ + 107 + ], + "ʒ": [ + 108 + ], + "ʔ": [ + 109 + ], + "ʕ": [ + 110 + ], + "ʘ": [ + 111 + ], + "ʙ": [ + 112 + ], + "ʛ": [ + 113 + ], + "ʜ": [ + 114 + ], + "ʝ": [ + 115 + ], + "ʟ": [ + 116 + ], + "ʡ": [ + 117 + ], + "ʢ": [ + 118 + ], + "ʲ": [ + 119 + ], + "ˈ": [ + 120 + ], + "ˌ": [ + 121 + ], + "ː": [ + 122 + ], + "ˑ": [ + 123 + ], + "˞": [ + 124 + ], + "β": [ + 125 + ], + "θ": [ + 126 + ], + "χ": [ + 127 + ], + "ᵻ": [ + 128 + ], + "ⱱ": [ + 129 + ], + "0": [ + 130 + ], + "1": [ + 131 + ], + "2": [ + 132 + ], + "3": [ + 133 + ], + "4": [ + 134 + ], + "5": [ + 135 + ], + "6": [ + 136 + ], + "7": [ + 137 + ], + "8": [ + 138 + ], + "9": [ + 139 + ], + "̧": [ + 140 + ], + "̃": [ + 141 + ], + "̪": [ + 142 + ], + "̯": [ + 143 + ], + "̩": [ + 144 + ], + "ʰ": [ + 145 + ], + "ˤ": [ + 146 + ], + "ε": [ + 147 + ], + "↓": [ + 148 + ], + "#": [ + 149 + ], + "\"": [ + 150 + ], + "↑": [ + 151 + ] + }, + "num_symbols": 256, + "num_speakers": 1, + "speaker_id_map": {}, + "piper_version": "1.0.0", + "language": { + "code": "pl_PL", + "family": "pl", + "region": "PL", + "name_native": "Polski", + "name_english": "Polish", + "country_english": "Poland" + }, + "dataset": "gosia" +} \ No newline at end of file diff --git a/pl/pl_PL/gosia/medium/samples/speaker_0.mp3 b/pl/pl_PL/gosia/medium/samples/speaker_0.mp3 new file mode 100644 index 0000000000000000000000000000000000000000..fbf93e3d2012aaa6c30cb2d175c298e5dc9ffd11 Binary files /dev/null and b/pl/pl_PL/gosia/medium/samples/speaker_0.mp3 differ diff --git a/pl/pl_PL/mc_speech/medium/MODEL_CARD b/pl/pl_PL/mc_speech/medium/MODEL_CARD new file mode 100644 index 0000000000000000000000000000000000000000..b8a34135d9764d857c2b8f7743870cfa81280d94 --- /dev/null +++ b/pl/pl_PL/mc_speech/medium/MODEL_CARD @@ -0,0 +1,15 @@ +# Model card for mc_speech (medium) + +* Language: pl_PL (Polish, Poland) +* Speakers: 1 +* Quality: medium +* Samplerate: 22,050Hz + +## Dataset + +* URL: https://www.kaggle.com/datasets/czyzi0/the-mc-speech-dataset +* License: CC0 + +## Training + +Finetuned from U.S. English lessac voice (medium quality). diff --git a/pl/pl_PL/mc_speech/medium/pl_PL-mc_speech-medium.onnx b/pl/pl_PL/mc_speech/medium/pl_PL-mc_speech-medium.onnx new file mode 100644 index 0000000000000000000000000000000000000000..11a723c0fa218cbbb4f9a2c212b0e7f69cdb86cc --- /dev/null +++ b/pl/pl_PL/mc_speech/medium/pl_PL-mc_speech-medium.onnx @@ -0,0 +1,3 @@ +version https://git-lfs.github.com/spec/v1 +oid sha256:a6b043358bc81e6c111a5140606a21959ce7f34969b8b7207f62869787cc3907 +size 63201294 diff --git a/pl/pl_PL/mc_speech/medium/pl_PL-mc_speech-medium.onnx.json b/pl/pl_PL/mc_speech/medium/pl_PL-mc_speech-medium.onnx.json new file mode 100644 index 0000000000000000000000000000000000000000..7fe77bf4481f5bed5052af31731f31f789183f05 --- /dev/null +++ b/pl/pl_PL/mc_speech/medium/pl_PL-mc_speech-medium.onnx.json @@ -0,0 +1,502 @@ +{ + "audio": { + "sample_rate": 22050, + "quality": "medium" + }, + "espeak": { + "voice": "pl" + }, + "inference": { + "noise_scale": 0.667, + "length_scale": 1, + "noise_w": 0.8 + }, + "phoneme_type": "espeak", + "phoneme_map": {}, + "phoneme_id_map": { + " ": [ + 3 + ], + "!": [ + 4 + ], + "\"": [ + 150 + ], + "#": [ + 149 + ], + "$": [ + 2 + ], + "'": [ + 5 + ], + "(": [ + 6 + ], + ")": [ + 7 + ], + ",": [ + 8 + ], + "-": [ + 9 + ], + ".": [ + 10 + ], + "0": [ + 130 + ], + "1": [ + 131 + ], + "2": [ + 132 + ], + "3": [ + 133 + ], + "4": [ + 134 + ], + "5": [ + 135 + ], + "6": [ + 136 + ], + "7": [ + 137 + ], + "8": [ + 138 + ], + "9": [ + 139 + ], + ":": [ + 11 + ], + ";": [ + 12 + ], + "?": [ + 13 + ], + "X": [ + 156 + ], + "^": [ + 1 + ], + "_": [ + 0 + ], + "a": [ + 14 + ], + "b": [ + 15 + ], + "c": [ + 16 + ], + "d": [ + 17 + ], + "e": [ + 18 + ], + "f": [ + 19 + ], + "g": [ + 154 + ], + "h": [ + 20 + ], + "i": [ + 21 + ], + "j": [ + 22 + ], + "k": [ + 23 + ], + "l": [ + 24 + ], + "m": [ + 25 + ], + "n": [ + 26 + ], + "o": [ + 27 + ], + "p": [ + 28 + ], + "q": [ + 29 + ], + "r": [ + 30 + ], + "s": [ + 31 + ], + "t": [ + 32 + ], + "u": [ + 33 + ], + "v": [ + 34 + ], + "w": [ + 35 + ], + "x": [ + 36 + ], + "y": [ + 37 + ], + "z": [ + 38 + ], + "æ": [ + 39 + ], + "ç": [ + 40 + ], + "ð": [ + 41 + ], + "ø": [ + 42 + ], + "ħ": [ + 43 + ], + "ŋ": [ + 44 + ], + "œ": [ + 45 + ], + "ǀ": [ + 46 + ], + "ǁ": [ + 47 + ], + "ǂ": [ + 48 + ], + "ǃ": [ + 49 + ], + "ɐ": [ + 50 + ], + "ɑ": [ + 51 + ], + "ɒ": [ + 52 + ], + "ɓ": [ + 53 + ], + "ɔ": [ + 54 + ], + "ɕ": [ + 55 + ], + "ɖ": [ + 56 + ], + "ɗ": [ + 57 + ], + "ɘ": [ + 58 + ], + "ə": [ + 59 + ], + "ɚ": [ + 60 + ], + "ɛ": [ + 61 + ], + "ɜ": [ + 62 + ], + "ɞ": [ + 63 + ], + "ɟ": [ + 64 + ], + "ɠ": [ + 65 + ], + "ɡ": [ + 66 + ], + "ɢ": [ + 67 + ], + "ɣ": [ + 68 + ], + "ɤ": [ + 69 + ], + "ɥ": [ + 70 + ], + "ɦ": [ + 71 + ], + "ɧ": [ + 72 + ], + "ɨ": [ + 73 + ], + "ɪ": [ + 74 + ], + "ɫ": [ + 75 + ], + "ɬ": [ + 76 + ], + "ɭ": [ + 77 + ], + "ɮ": [ + 78 + ], + "ɯ": [ + 79 + ], + "ɰ": [ + 80 + ], + "ɱ": [ + 81 + ], + "ɲ": [ + 82 + ], + "ɳ": [ + 83 + ], + "ɴ": [ + 84 + ], + "ɵ": [ + 85 + ], + "ɶ": [ + 86 + ], + "ɸ": [ + 87 + ], + "ɹ": [ + 88 + ], + "ɺ": [ + 89 + ], + "ɻ": [ + 90 + ], + "ɽ": [ + 91 + ], + "ɾ": [ + 92 + ], + "ʀ": [ + 93 + ], + "ʁ": [ + 94 + ], + "ʂ": [ + 95 + ], + "ʃ": [ + 96 + ], + "ʄ": [ + 97 + ], + "ʈ": [ + 98 + ], + "ʉ": [ + 99 + ], + "ʊ": [ + 100 + ], + "ʋ": [ + 101 + ], + "ʌ": [ + 102 + ], + "ʍ": [ + 103 + ], + "ʎ": [ + 104 + ], + "ʏ": [ + 105 + ], + "ʐ": [ + 106 + ], + "ʑ": [ + 107 + ], + "ʒ": [ + 108 + ], + "ʔ": [ + 109 + ], + "ʕ": [ + 110 + ], + "ʘ": [ + 111 + ], + "ʙ": [ + 112 + ], + "ʛ": [ + 113 + ], + "ʜ": [ + 114 + ], + "ʝ": [ + 115 + ], + "ʟ": [ + 116 + ], + "ʡ": [ + 117 + ], + "ʢ": [ + 118 + ], + "ʦ": [ + 155 + ], + "ʰ": [ + 145 + ], + "ʲ": [ + 119 + ], + "ˈ": [ + 120 + ], + "ˌ": [ + 121 + ], + "ː": [ + 122 + ], + "ˑ": [ + 123 + ], + "˞": [ + 124 + ], + "ˤ": [ + 146 + ], + "̃": [ + 141 + ], + "̧": [ + 140 + ], + "̩": [ + 144 + ], + "̪": [ + 142 + ], + "̯": [ + 143 + ], + "̺": [ + 152 + ], + "̻": [ + 153 + ], + "β": [ + 125 + ], + "ε": [ + 147 + ], + "θ": [ + 126 + ], + "χ": [ + 127 + ], + "ᵻ": [ + 128 + ], + "↑": [ + 151 + ], + "↓": [ + 148 + ], + "ⱱ": [ + 129 + ] + }, + "num_symbols": 256, + "num_speakers": 1, + "speaker_id_map": {}, + "piper_version": "1.0.0", + "language": { + "code": "pl_PL", + "family": "pl", + "region": "PL", + "name_native": "Polski", + "name_english": "Polish", + "country_english": "Poland" + }, + "dataset": "mc_speech" +} \ No newline at end of file diff --git a/pl/pl_PL/mc_speech/medium/samples/speaker_0.mp3 b/pl/pl_PL/mc_speech/medium/samples/speaker_0.mp3 new file mode 100644 index 0000000000000000000000000000000000000000..91fce8d9b0d99e80f471bb83c3e9fd9e30321e26 Binary files /dev/null and b/pl/pl_PL/mc_speech/medium/samples/speaker_0.mp3 differ diff --git a/pl/pl_PL/mls_6892/low/ALIASES b/pl/pl_PL/mls_6892/low/ALIASES new file mode 100644 index 0000000000000000000000000000000000000000..99705771e57d94aed3f50cc4347b6e64a80b1974 --- /dev/null +++ b/pl/pl_PL/mls_6892/low/ALIASES @@ -0,0 +1 @@ +pl-mls_6892-low diff --git a/pl/pl_PL/mls_6892/low/MODEL_CARD b/pl/pl_PL/mls_6892/low/MODEL_CARD new file mode 100644 index 0000000000000000000000000000000000000000..aa92e602122dc99373fc677b91e77c7a9127ad1f --- /dev/null +++ b/pl/pl_PL/mls_6892/low/MODEL_CARD @@ -0,0 +1,15 @@ +# Model card for mls_6892 (low) + +* Language: pl_PL (Polish, Polish) +* Speakers: 1 +* Quality: low +* Samplerate: 16,000Hz + +## Dataset + +* URL: http://www.openslr.org/94/ +* License: CC BY 4.0 + +## Training + +Finetuned from U.S. English Ryan voice (low quality). diff --git a/pl/pl_PL/mls_6892/low/pl_PL-mls_6892-low.onnx b/pl/pl_PL/mls_6892/low/pl_PL-mls_6892-low.onnx new file mode 100644 index 0000000000000000000000000000000000000000..afe0168192a4b7d06c751a5527836a0b66bf0867 --- /dev/null +++ b/pl/pl_PL/mls_6892/low/pl_PL-mls_6892-low.onnx @@ -0,0 +1,3 @@ +version https://git-lfs.github.com/spec/v1 +oid sha256:b572ab1a0f1d3b722b2a545e0b38834e6f932852cd0f11a44c022baa8f104c10 +size 63104526 diff --git a/pl/pl_PL/mls_6892/low/pl_PL-mls_6892-low.onnx.json b/pl/pl_PL/mls_6892/low/pl_PL-mls_6892-low.onnx.json new file mode 100644 index 0000000000000000000000000000000000000000..46f02d24b646a870ced2cb1891a5206d63ae1cba --- /dev/null +++ b/pl/pl_PL/mls_6892/low/pl_PL-mls_6892-low.onnx.json @@ -0,0 +1,420 @@ +{ + "audio": { + "sample_rate": 16000, + "quality": "low" + }, + "espeak": { + "voice": "pl" + }, + "inference": { + "noise_scale": 0.667, + "length_scale": 1, + "noise_w": 0.8 + }, + "phoneme_map": {}, + "phoneme_id_map": { + "_": [ + 0 + ], + "^": [ + 1 + ], + "$": [ + 2 + ], + " ": [ + 3 + ], + "!": [ + 4 + ], + "'": [ + 5 + ], + "(": [ + 6 + ], + ")": [ + 7 + ], + ",": [ + 8 + ], + "-": [ + 9 + ], + ".": [ + 10 + ], + ":": [ + 11 + ], + ";": [ + 12 + ], + "?": [ + 13 + ], + "a": [ + 14 + ], + "b": [ + 15 + ], + "c": [ + 16 + ], + "d": [ + 17 + ], + "e": [ + 18 + ], + "f": [ + 19 + ], + "h": [ + 20 + ], + "i": [ + 21 + ], + "j": [ + 22 + ], + "k": [ + 23 + ], + "l": [ + 24 + ], + "m": [ + 25 + ], + "n": [ + 26 + ], + "o": [ + 27 + ], + "p": [ + 28 + ], + "q": [ + 29 + ], + "r": [ + 30 + ], + "s": [ + 31 + ], + "t": [ + 32 + ], + "u": [ + 33 + ], + "v": [ + 34 + ], + "w": [ + 35 + ], + "x": [ + 36 + ], + "y": [ + 37 + ], + "z": [ + 38 + ], + "æ": [ + 39 + ], + "ç": [ + 40 + ], + "ð": [ + 41 + ], + "ø": [ + 42 + ], + "ħ": [ + 43 + ], + "ŋ": [ + 44 + ], + "œ": [ + 45 + ], + "ǀ": [ + 46 + ], + "ǁ": [ + 47 + ], + "ǂ": [ + 48 + ], + "ǃ": [ + 49 + ], + "ɐ": [ + 50 + ], + "ɑ": [ + 51 + ], + "ɒ": [ + 52 + ], + "ɓ": [ + 53 + ], + "ɔ": [ + 54 + ], + "ɕ": [ + 55 + ], + "ɖ": [ + 56 + ], + "ɗ": [ + 57 + ], + "ɘ": [ + 58 + ], + "ə": [ + 59 + ], + "ɚ": [ + 60 + ], + "ɛ": [ + 61 + ], + "ɜ": [ + 62 + ], + "ɞ": [ + 63 + ], + "ɟ": [ + 64 + ], + "ɠ": [ + 65 + ], + "ɡ": [ + 66 + ], + "ɢ": [ + 67 + ], + "ɣ": [ + 68 + ], + "ɤ": [ + 69 + ], + "ɥ": [ + 70 + ], + "ɦ": [ + 71 + ], + "ɧ": [ + 72 + ], + "ɨ": [ + 73 + ], + "ɪ": [ + 74 + ], + "ɫ": [ + 75 + ], + "ɬ": [ + 76 + ], + "ɭ": [ + 77 + ], + "ɮ": [ + 78 + ], + "ɯ": [ + 79 + ], + "ɰ": [ + 80 + ], + "ɱ": [ + 81 + ], + "ɲ": [ + 82 + ], + "ɳ": [ + 83 + ], + "ɴ": [ + 84 + ], + "ɵ": [ + 85 + ], + "ɶ": [ + 86 + ], + "ɸ": [ + 87 + ], + "ɹ": [ + 88 + ], + "ɺ": [ + 89 + ], + "ɻ": [ + 90 + ], + "ɽ": [ + 91 + ], + "ɾ": [ + 92 + ], + "ʀ": [ + 93 + ], + "ʁ": [ + 94 + ], + "ʂ": [ + 95 + ], + "ʃ": [ + 96 + ], + "ʄ": [ + 97 + ], + "ʈ": [ + 98 + ], + "ʉ": [ + 99 + ], + "ʊ": [ + 100 + ], + "ʋ": [ + 101 + ], + "ʌ": [ + 102 + ], + "ʍ": [ + 103 + ], + "ʎ": [ + 104 + ], + "ʏ": [ + 105 + ], + "ʐ": [ + 106 + ], + "ʑ": [ + 107 + ], + "ʒ": [ + 108 + ], + "ʔ": [ + 109 + ], + "ʕ": [ + 110 + ], + "ʘ": [ + 111 + ], + "ʙ": [ + 112 + ], + "ʛ": [ + 113 + ], + "ʜ": [ + 114 + ], + "ʝ": [ + 115 + ], + "ʟ": [ + 116 + ], + "ʡ": [ + 117 + ], + "ʢ": [ + 118 + ], + "ʲ": [ + 119 + ], + "ˈ": [ + 120 + ], + "ˌ": [ + 121 + ], + "ː": [ + 122 + ], + "ˑ": [ + 123 + ], + "˞": [ + 124 + ], + "β": [ + 125 + ], + "θ": [ + 126 + ], + "χ": [ + 127 + ], + "ᵻ": [ + 128 + ], + "ⱱ": [ + 129 + ] + }, + "num_symbols": 130, + "num_speakers": 1, + "speaker_id_map": {}, + "piper_version": "0.2.0", + "language": { + "code": "pl_PL", + "family": "pl", + "region": "PL", + "name_native": "Polski", + "name_english": "Polish", + "country_english": "Poland" + }, + "dataset": "mls_6892" +} \ No newline at end of file diff --git a/pl/pl_PL/mls_6892/low/samples/speaker_0.mp3 b/pl/pl_PL/mls_6892/low/samples/speaker_0.mp3 new file mode 100644 index 0000000000000000000000000000000000000000..bbabe91d1691632df6d71abe5ba2b6190e17f6d1 Binary files /dev/null and b/pl/pl_PL/mls_6892/low/samples/speaker_0.mp3 differ diff --git a/pt/pt_BR/edresson/low/ALIASES b/pt/pt_BR/edresson/low/ALIASES new file mode 100644 index 0000000000000000000000000000000000000000..1e7e27917328d7e3287287b07af95897c12a6b0e --- /dev/null +++ b/pt/pt_BR/edresson/low/ALIASES @@ -0,0 +1 @@ +pt-br-edresson-low diff --git a/pt/pt_BR/edresson/low/MODEL_CARD b/pt/pt_BR/edresson/low/MODEL_CARD new file mode 100644 index 0000000000000000000000000000000000000000..0c11733ca3e9bc68d2b754eb2dfaaad5a03730d1 --- /dev/null +++ b/pt/pt_BR/edresson/low/MODEL_CARD @@ -0,0 +1,15 @@ +# Model card for edresson (low) + +* Language: pt_BR (Portuguese, Brazil) +* Speakers: 1 +* Quality: low +* Samplerate: 16,000Hz + +## Dataset + +* URL: https://github.com/Edresson/TTS-Portuguese-Corpus +* License: CC BY 4.0 + +## Training + +Finetuned from U.S. English Ryan voice (low quality). diff --git a/pt/pt_BR/edresson/low/pt_BR-edresson-low.onnx b/pt/pt_BR/edresson/low/pt_BR-edresson-low.onnx new file mode 100644 index 0000000000000000000000000000000000000000..d7807968b6dae66e14eea4dfacdb46683f548327 --- /dev/null +++ b/pt/pt_BR/edresson/low/pt_BR-edresson-low.onnx @@ -0,0 +1,3 @@ +version https://git-lfs.github.com/spec/v1 +oid sha256:de4cecee38b30bb1a6378a337af605d59f0c377df702c6a6752870db8991cd84 +size 63104526 diff --git a/pt/pt_BR/edresson/low/pt_BR-edresson-low.onnx.json b/pt/pt_BR/edresson/low/pt_BR-edresson-low.onnx.json new file mode 100644 index 0000000000000000000000000000000000000000..43925b4324e3b25a96298d4b8a7d907ac9867937 --- /dev/null +++ b/pt/pt_BR/edresson/low/pt_BR-edresson-low.onnx.json @@ -0,0 +1,420 @@ +{ + "audio": { + "sample_rate": 16000, + "quality": "low" + }, + "espeak": { + "voice": "pt-br" + }, + "inference": { + "noise_scale": 0.667, + "length_scale": 1, + "noise_w": 0.8 + }, + "phoneme_map": {}, + "phoneme_id_map": { + "_": [ + 0 + ], + "^": [ + 1 + ], + "$": [ + 2 + ], + " ": [ + 3 + ], + "!": [ + 4 + ], + "'": [ + 5 + ], + "(": [ + 6 + ], + ")": [ + 7 + ], + ",": [ + 8 + ], + "-": [ + 9 + ], + ".": [ + 10 + ], + ":": [ + 11 + ], + ";": [ + 12 + ], + "?": [ + 13 + ], + "a": [ + 14 + ], + "b": [ + 15 + ], + "c": [ + 16 + ], + "d": [ + 17 + ], + "e": [ + 18 + ], + "f": [ + 19 + ], + "h": [ + 20 + ], + "i": [ + 21 + ], + "j": [ + 22 + ], + "k": [ + 23 + ], + "l": [ + 24 + ], + "m": [ + 25 + ], + "n": [ + 26 + ], + "o": [ + 27 + ], + "p": [ + 28 + ], + "q": [ + 29 + ], + "r": [ + 30 + ], + "s": [ + 31 + ], + "t": [ + 32 + ], + "u": [ + 33 + ], + "v": [ + 34 + ], + "w": [ + 35 + ], + "x": [ + 36 + ], + "y": [ + 37 + ], + "z": [ + 38 + ], + "æ": [ + 39 + ], + "ç": [ + 40 + ], + "ð": [ + 41 + ], + "ø": [ + 42 + ], + "ħ": [ + 43 + ], + "ŋ": [ + 44 + ], + "œ": [ + 45 + ], + "ǀ": [ + 46 + ], + "ǁ": [ + 47 + ], + "ǂ": [ + 48 + ], + "ǃ": [ + 49 + ], + "ɐ": [ + 50 + ], + "ɑ": [ + 51 + ], + "ɒ": [ + 52 + ], + "ɓ": [ + 53 + ], + "ɔ": [ + 54 + ], + "ɕ": [ + 55 + ], + "ɖ": [ + 56 + ], + "ɗ": [ + 57 + ], + "ɘ": [ + 58 + ], + "ə": [ + 59 + ], + "ɚ": [ + 60 + ], + "ɛ": [ + 61 + ], + "ɜ": [ + 62 + ], + "ɞ": [ + 63 + ], + "ɟ": [ + 64 + ], + "ɠ": [ + 65 + ], + "ɡ": [ + 66 + ], + "ɢ": [ + 67 + ], + "ɣ": [ + 68 + ], + "ɤ": [ + 69 + ], + "ɥ": [ + 70 + ], + "ɦ": [ + 71 + ], + "ɧ": [ + 72 + ], + "ɨ": [ + 73 + ], + "ɪ": [ + 74 + ], + "ɫ": [ + 75 + ], + "ɬ": [ + 76 + ], + "ɭ": [ + 77 + ], + "ɮ": [ + 78 + ], + "ɯ": [ + 79 + ], + "ɰ": [ + 80 + ], + "ɱ": [ + 81 + ], + "ɲ": [ + 82 + ], + "ɳ": [ + 83 + ], + "ɴ": [ + 84 + ], + "ɵ": [ + 85 + ], + "ɶ": [ + 86 + ], + "ɸ": [ + 87 + ], + "ɹ": [ + 88 + ], + "ɺ": [ + 89 + ], + "ɻ": [ + 90 + ], + "ɽ": [ + 91 + ], + "ɾ": [ + 92 + ], + "ʀ": [ + 93 + ], + "ʁ": [ + 94 + ], + "ʂ": [ + 95 + ], + "ʃ": [ + 96 + ], + "ʄ": [ + 97 + ], + "ʈ": [ + 98 + ], + "ʉ": [ + 99 + ], + "ʊ": [ + 100 + ], + "ʋ": [ + 101 + ], + "ʌ": [ + 102 + ], + "ʍ": [ + 103 + ], + "ʎ": [ + 104 + ], + "ʏ": [ + 105 + ], + "ʐ": [ + 106 + ], + "ʑ": [ + 107 + ], + "ʒ": [ + 108 + ], + "ʔ": [ + 109 + ], + "ʕ": [ + 110 + ], + "ʘ": [ + 111 + ], + "ʙ": [ + 112 + ], + "ʛ": [ + 113 + ], + "ʜ": [ + 114 + ], + "ʝ": [ + 115 + ], + "ʟ": [ + 116 + ], + "ʡ": [ + 117 + ], + "ʢ": [ + 118 + ], + "ʲ": [ + 119 + ], + "ˈ": [ + 120 + ], + "ˌ": [ + 121 + ], + "ː": [ + 122 + ], + "ˑ": [ + 123 + ], + "˞": [ + 124 + ], + "β": [ + 125 + ], + "θ": [ + 126 + ], + "χ": [ + 127 + ], + "ᵻ": [ + 128 + ], + "ⱱ": [ + 129 + ] + }, + "num_symbols": 130, + "num_speakers": 1, + "speaker_id_map": {}, + "piper_version": "0.2.0", + "language": { + "code": "pt_BR", + "family": "pt", + "region": "BR", + "name_native": "Português", + "name_english": "Portuguese", + "country_english": "Brazil" + }, + "dataset": "edresson" +} \ No newline at end of file diff --git a/pt/pt_BR/edresson/low/samples/speaker_0.mp3 b/pt/pt_BR/edresson/low/samples/speaker_0.mp3 new file mode 100644 index 0000000000000000000000000000000000000000..0f8fec664efdb0382daba72140133229fb085879 Binary files /dev/null and b/pt/pt_BR/edresson/low/samples/speaker_0.mp3 differ diff --git a/pt/pt_BR/faber/medium/MODEL_CARD b/pt/pt_BR/faber/medium/MODEL_CARD new file mode 100644 index 0000000000000000000000000000000000000000..d20a0a0d7502c64f56c11d1becd2e5bdf3916fa2 --- /dev/null +++ b/pt/pt_BR/faber/medium/MODEL_CARD @@ -0,0 +1,15 @@ +# Model card for faber (medium) + +* Language: pt_BR (Portuguese, Brazil) +* Speakers: 1 +* Quality: medium +* Samplerate: 22,050Hz + +## Dataset + +* URL: https://github.com/NabuCasa/voice-datasets +* License: CC0 + +## Training + +Finetuned from U.S. English lessac voice (medium quality). diff --git a/pt/pt_BR/faber/medium/pt_BR-faber-medium.onnx b/pt/pt_BR/faber/medium/pt_BR-faber-medium.onnx new file mode 100644 index 0000000000000000000000000000000000000000..324a5f23579ef280c7aa0f5c32a0ac301a2f765a --- /dev/null +++ b/pt/pt_BR/faber/medium/pt_BR-faber-medium.onnx @@ -0,0 +1,3 @@ +version https://git-lfs.github.com/spec/v1 +oid sha256:858555e3a064209c57088fe6bd70c4c3dc54d03eaa00c45d5ecaf43a33f95aa7 +size 63201294 diff --git a/pt/pt_BR/faber/medium/pt_BR-faber-medium.onnx.json b/pt/pt_BR/faber/medium/pt_BR-faber-medium.onnx.json new file mode 100644 index 0000000000000000000000000000000000000000..b0206c81d70e6e4468b94d36491b4c17fac531b7 --- /dev/null +++ b/pt/pt_BR/faber/medium/pt_BR-faber-medium.onnx.json @@ -0,0 +1,491 @@ +{ + "audio": { + "sample_rate": 22050, + "quality": "medium" + }, + "espeak": { + "voice": "pt-br" + }, + "inference": { + "noise_scale": 0.667, + "length_scale": 1, + "noise_w": 0.8 + }, + "phoneme_type": "espeak", + "phoneme_map": { + "c": [ + "k" + ] + }, + "phoneme_id_map": { + "_": [ + 0 + ], + "^": [ + 1 + ], + "$": [ + 2 + ], + " ": [ + 3 + ], + "!": [ + 4 + ], + "'": [ + 5 + ], + "(": [ + 6 + ], + ")": [ + 7 + ], + ",": [ + 8 + ], + "-": [ + 9 + ], + ".": [ + 10 + ], + ":": [ + 11 + ], + ";": [ + 12 + ], + "?": [ + 13 + ], + "a": [ + 14 + ], + "b": [ + 15 + ], + "c": [ + 16 + ], + "d": [ + 17 + ], + "e": [ + 18 + ], + "f": [ + 19 + ], + "h": [ + 20 + ], + "i": [ + 21 + ], + "j": [ + 22 + ], + "k": [ + 23 + ], + "l": [ + 24 + ], + "m": [ + 25 + ], + "n": [ + 26 + ], + "o": [ + 27 + ], + "p": [ + 28 + ], + "q": [ + 29 + ], + "r": [ + 30 + ], + "s": [ + 31 + ], + "t": [ + 32 + ], + "u": [ + 33 + ], + "v": [ + 34 + ], + "w": [ + 35 + ], + "x": [ + 36 + ], + "y": [ + 37 + ], + "z": [ + 38 + ], + "æ": [ + 39 + ], + "ç": [ + 40 + ], + "ð": [ + 41 + ], + "ø": [ + 42 + ], + "ħ": [ + 43 + ], + "ŋ": [ + 44 + ], + "œ": [ + 45 + ], + "ǀ": [ + 46 + ], + "ǁ": [ + 47 + ], + "ǂ": [ + 48 + ], + "ǃ": [ + 49 + ], + "ɐ": [ + 50 + ], + "ɑ": [ + 51 + ], + "ɒ": [ + 52 + ], + "ɓ": [ + 53 + ], + "ɔ": [ + 54 + ], + "ɕ": [ + 55 + ], + "ɖ": [ + 56 + ], + "ɗ": [ + 57 + ], + "ɘ": [ + 58 + ], + "ə": [ + 59 + ], + "ɚ": [ + 60 + ], + "ɛ": [ + 61 + ], + "ɜ": [ + 62 + ], + "ɞ": [ + 63 + ], + "ɟ": [ + 64 + ], + "ɠ": [ + 65 + ], + "ɡ": [ + 66 + ], + "ɢ": [ + 67 + ], + "ɣ": [ + 68 + ], + "ɤ": [ + 69 + ], + "ɥ": [ + 70 + ], + "ɦ": [ + 71 + ], + "ɧ": [ + 72 + ], + "ɨ": [ + 73 + ], + "ɪ": [ + 74 + ], + "ɫ": [ + 75 + ], + "ɬ": [ + 76 + ], + "ɭ": [ + 77 + ], + "ɮ": [ + 78 + ], + "ɯ": [ + 79 + ], + "ɰ": [ + 80 + ], + "ɱ": [ + 81 + ], + "ɲ": [ + 82 + ], + "ɳ": [ + 83 + ], + "ɴ": [ + 84 + ], + "ɵ": [ + 85 + ], + "ɶ": [ + 86 + ], + "ɸ": [ + 87 + ], + "ɹ": [ + 88 + ], + "ɺ": [ + 89 + ], + "ɻ": [ + 90 + ], + "ɽ": [ + 91 + ], + "ɾ": [ + 92 + ], + "ʀ": [ + 93 + ], + "ʁ": [ + 94 + ], + "ʂ": [ + 95 + ], + "ʃ": [ + 96 + ], + "ʄ": [ + 97 + ], + "ʈ": [ + 98 + ], + "ʉ": [ + 99 + ], + "ʊ": [ + 100 + ], + "ʋ": [ + 101 + ], + "ʌ": [ + 102 + ], + "ʍ": [ + 103 + ], + "ʎ": [ + 104 + ], + "ʏ": [ + 105 + ], + "ʐ": [ + 106 + ], + "ʑ": [ + 107 + ], + "ʒ": [ + 108 + ], + "ʔ": [ + 109 + ], + "ʕ": [ + 110 + ], + "ʘ": [ + 111 + ], + "ʙ": [ + 112 + ], + "ʛ": [ + 113 + ], + "ʜ": [ + 114 + ], + "ʝ": [ + 115 + ], + "ʟ": [ + 116 + ], + "ʡ": [ + 117 + ], + "ʢ": [ + 118 + ], + "ʲ": [ + 119 + ], + "ˈ": [ + 120 + ], + "ˌ": [ + 121 + ], + "ː": [ + 122 + ], + "ˑ": [ + 123 + ], + "˞": [ + 124 + ], + "β": [ + 125 + ], + "θ": [ + 126 + ], + "χ": [ + 127 + ], + "ᵻ": [ + 128 + ], + "ⱱ": [ + 129 + ], + "0": [ + 130 + ], + "1": [ + 131 + ], + "2": [ + 132 + ], + "3": [ + 133 + ], + "4": [ + 134 + ], + "5": [ + 135 + ], + "6": [ + 136 + ], + "7": [ + 137 + ], + "8": [ + 138 + ], + "9": [ + 139 + ], + "̧": [ + 140 + ], + "̃": [ + 141 + ], + "̪": [ + 142 + ], + "̯": [ + 143 + ], + "̩": [ + 144 + ], + "ʰ": [ + 145 + ], + "ˤ": [ + 146 + ], + "ε": [ + 147 + ], + "↓": [ + 148 + ], + "#": [ + 149 + ], + "\"": [ + 150 + ], + "↑": [ + 151 + ] + }, + "num_symbols": 256, + "num_speakers": 1, + "speaker_id_map": {}, + "piper_version": "1.0.0", + "language": { + "code": "pt_BR", + "family": "pt", + "region": "BR", + "name_native": "Português", + "name_english": "Portuguese", + "country_english": "Brazil" + }, + "dataset": "faber" +} \ No newline at end of file diff --git a/pt/pt_BR/faber/medium/samples/speaker_0.mp3 b/pt/pt_BR/faber/medium/samples/speaker_0.mp3 new file mode 100644 index 0000000000000000000000000000000000000000..292e7f694ec2a7a0fdcd82008f126f33c828f295 Binary files /dev/null and b/pt/pt_BR/faber/medium/samples/speaker_0.mp3 differ diff --git "a/pt/pt_PT/tug\303\243o/medium/MODEL_CARD" "b/pt/pt_PT/tug\303\243o/medium/MODEL_CARD" new file mode 100644 index 0000000000000000000000000000000000000000..84d864ab44ac95aa72abc92a9ad2a3e3a8aa2e89 --- /dev/null +++ "b/pt/pt_PT/tug\303\243o/medium/MODEL_CARD" @@ -0,0 +1,15 @@ +# Model card for tugão (medium) + +* Language: pt_PT (Portuguese, Portugal) +* Speakers: 1 +* Quality: medium +* Samplerate: 22,050Hz + +## Dataset + +* URL: https://github.com/NabuCasa/voice-datasets +* License: CC0 + +## Training + +Finetuned from U.S. English lessac voice (medium quality). diff --git "a/pt/pt_PT/tug\303\243o/medium/pt_PT-tug\303\243o-medium.onnx" "b/pt/pt_PT/tug\303\243o/medium/pt_PT-tug\303\243o-medium.onnx" new file mode 100644 index 0000000000000000000000000000000000000000..217248f2a203b9840f7db094a314b538da7daf3f --- /dev/null +++ "b/pt/pt_PT/tug\303\243o/medium/pt_PT-tug\303\243o-medium.onnx" @@ -0,0 +1,3 @@ +version https://git-lfs.github.com/spec/v1 +oid sha256:223a7aaca69a155c61897e8ada7c3b13bc306e16c72dbb9c2fed733e2b0927d4 +size 63201294 diff --git "a/pt/pt_PT/tug\303\243o/medium/pt_PT-tug\303\243o-medium.onnx.json" "b/pt/pt_PT/tug\303\243o/medium/pt_PT-tug\303\243o-medium.onnx.json" new file mode 100644 index 0000000000000000000000000000000000000000..445c1393ec16e8a0fb9687592cbd883f4fd6bd88 --- /dev/null +++ "b/pt/pt_PT/tug\303\243o/medium/pt_PT-tug\303\243o-medium.onnx.json" @@ -0,0 +1,508 @@ +{ + "audio": { + "sample_rate": 22050, + "quality": "medium" + }, + "espeak": { + "voice": "pt" + }, + "inference": { + "noise_scale": 0.667, + "length_scale": 1, + "noise_w": 0.8 + }, + "phoneme_type": "espeak", + "phoneme_map": {}, + "phoneme_id_map": { + " ": [ + 3 + ], + "!": [ + 4 + ], + "\"": [ + 150 + ], + "#": [ + 149 + ], + "$": [ + 2 + ], + "'": [ + 5 + ], + "(": [ + 6 + ], + ")": [ + 7 + ], + ",": [ + 8 + ], + "-": [ + 9 + ], + ".": [ + 10 + ], + "0": [ + 130 + ], + "1": [ + 131 + ], + "2": [ + 132 + ], + "3": [ + 133 + ], + "4": [ + 134 + ], + "5": [ + 135 + ], + "6": [ + 136 + ], + "7": [ + 137 + ], + "8": [ + 138 + ], + "9": [ + 139 + ], + ":": [ + 11 + ], + ";": [ + 12 + ], + "?": [ + 13 + ], + "X": [ + 156 + ], + "^": [ + 1 + ], + "_": [ + 0 + ], + "a": [ + 14 + ], + "b": [ + 15 + ], + "c": [ + 16 + ], + "d": [ + 17 + ], + "e": [ + 18 + ], + "f": [ + 19 + ], + "g": [ + 154 + ], + "h": [ + 20 + ], + "i": [ + 21 + ], + "j": [ + 22 + ], + "k": [ + 23 + ], + "l": [ + 24 + ], + "m": [ + 25 + ], + "n": [ + 26 + ], + "o": [ + 27 + ], + "p": [ + 28 + ], + "q": [ + 29 + ], + "r": [ + 30 + ], + "s": [ + 31 + ], + "t": [ + 32 + ], + "u": [ + 33 + ], + "v": [ + 34 + ], + "w": [ + 35 + ], + "x": [ + 36 + ], + "y": [ + 37 + ], + "z": [ + 38 + ], + "æ": [ + 39 + ], + "ç": [ + 40 + ], + "ð": [ + 41 + ], + "ø": [ + 42 + ], + "ħ": [ + 43 + ], + "ŋ": [ + 44 + ], + "œ": [ + 45 + ], + "ǀ": [ + 46 + ], + "ǁ": [ + 47 + ], + "ǂ": [ + 48 + ], + "ǃ": [ + 49 + ], + "ɐ": [ + 50 + ], + "ɑ": [ + 51 + ], + "ɒ": [ + 52 + ], + "ɓ": [ + 53 + ], + "ɔ": [ + 54 + ], + "ɕ": [ + 55 + ], + "ɖ": [ + 56 + ], + "ɗ": [ + 57 + ], + "ɘ": [ + 58 + ], + "ə": [ + 59 + ], + "ɚ": [ + 60 + ], + "ɛ": [ + 61 + ], + "ɜ": [ + 62 + ], + "ɞ": [ + 63 + ], + "ɟ": [ + 64 + ], + "ɠ": [ + 65 + ], + "ɡ": [ + 66 + ], + "ɢ": [ + 67 + ], + "ɣ": [ + 68 + ], + "ɤ": [ + 69 + ], + "ɥ": [ + 70 + ], + "ɦ": [ + 71 + ], + "ɧ": [ + 72 + ], + "ɨ": [ + 73 + ], + "ɪ": [ + 74 + ], + "ɫ": [ + 75 + ], + "ɬ": [ + 76 + ], + "ɭ": [ + 77 + ], + "ɮ": [ + 78 + ], + "ɯ": [ + 79 + ], + "ɰ": [ + 80 + ], + "ɱ": [ + 81 + ], + "ɲ": [ + 82 + ], + "ɳ": [ + 83 + ], + "ɴ": [ + 84 + ], + "ɵ": [ + 85 + ], + "ɶ": [ + 86 + ], + "ɸ": [ + 87 + ], + "ɹ": [ + 88 + ], + "ɺ": [ + 89 + ], + "ɻ": [ + 90 + ], + "ɽ": [ + 91 + ], + "ɾ": [ + 92 + ], + "ʀ": [ + 93 + ], + "ʁ": [ + 94 + ], + "ʂ": [ + 95 + ], + "ʃ": [ + 96 + ], + "ʄ": [ + 97 + ], + "ʈ": [ + 98 + ], + "ʉ": [ + 99 + ], + "ʊ": [ + 100 + ], + "ʋ": [ + 101 + ], + "ʌ": [ + 102 + ], + "ʍ": [ + 103 + ], + "ʎ": [ + 104 + ], + "ʏ": [ + 105 + ], + "ʐ": [ + 106 + ], + "ʑ": [ + 107 + ], + "ʒ": [ + 108 + ], + "ʔ": [ + 109 + ], + "ʕ": [ + 110 + ], + "ʘ": [ + 111 + ], + "ʙ": [ + 112 + ], + "ʛ": [ + 113 + ], + "ʜ": [ + 114 + ], + "ʝ": [ + 115 + ], + "ʟ": [ + 116 + ], + "ʡ": [ + 117 + ], + "ʢ": [ + 118 + ], + "ʦ": [ + 155 + ], + "ʰ": [ + 145 + ], + "ʲ": [ + 119 + ], + "ˈ": [ + 120 + ], + "ˌ": [ + 121 + ], + "ː": [ + 122 + ], + "ˑ": [ + 123 + ], + "˞": [ + 124 + ], + "ˤ": [ + 146 + ], + "̃": [ + 141 + ], + "̊": [ + 158 + ], + "̝": [ + 157 + ], + "̧": [ + 140 + ], + "̩": [ + 144 + ], + "̪": [ + 142 + ], + "̯": [ + 143 + ], + "̺": [ + 152 + ], + "̻": [ + 153 + ], + "β": [ + 125 + ], + "ε": [ + 147 + ], + "θ": [ + 126 + ], + "χ": [ + 127 + ], + "ᵻ": [ + 128 + ], + "↑": [ + 151 + ], + "↓": [ + 148 + ], + "ⱱ": [ + 129 + ] + }, + "num_symbols": 256, + "num_speakers": 1, + "speaker_id_map": {}, + "piper_version": "1.0.0", + "language": { + "code": "pt_PT", + "family": "pt", + "region": "PT", + "name_native": "Português", + "name_english": "Portuguese", + "country_english": "Portugal" + }, + "dataset": "tugão" +} \ No newline at end of file diff --git "a/pt/pt_PT/tug\303\243o/medium/samples/speaker_0.mp3" "b/pt/pt_PT/tug\303\243o/medium/samples/speaker_0.mp3" new file mode 100644 index 0000000000000000000000000000000000000000..9ff347c35430f9a222018695c909a568e15544fa Binary files /dev/null and "b/pt/pt_PT/tug\303\243o/medium/samples/speaker_0.mp3" differ diff --git a/ro/ro_RO/mihai/medium/MODEL_CARD b/ro/ro_RO/mihai/medium/MODEL_CARD new file mode 100644 index 0000000000000000000000000000000000000000..abef88293e6eef28f3571250334de87dba756fd2 --- /dev/null +++ b/ro/ro_RO/mihai/medium/MODEL_CARD @@ -0,0 +1,15 @@ +# Model card for miahi (medium) + +* Language: ro_RO (Romanian, Romania) +* Speakers: 1 +* Quality: medium +* Samplerate: 22,050Hz + +## Dataset + +* URL: https://github.com/NabuCasa/voice-datasets +* License: CC0 + +## Training + +Finetuned from U.S. English lessac voice (medium quality). diff --git a/ro/ro_RO/mihai/medium/ro_RO-mihai-medium.onnx b/ro/ro_RO/mihai/medium/ro_RO-mihai-medium.onnx new file mode 100644 index 0000000000000000000000000000000000000000..f8d3dc617ffa31b16e0a2e1b90f972d0ce1829b4 --- /dev/null +++ b/ro/ro_RO/mihai/medium/ro_RO-mihai-medium.onnx @@ -0,0 +1,3 @@ +version https://git-lfs.github.com/spec/v1 +oid sha256:e0608bbbd53c80267c09ece681b09f5199f54e792356684c8073738e5f15d29f +size 63201294 diff --git a/ro/ro_RO/mihai/medium/ro_RO-mihai-medium.onnx.json b/ro/ro_RO/mihai/medium/ro_RO-mihai-medium.onnx.json new file mode 100644 index 0000000000000000000000000000000000000000..f28756abb578bac29f0f6742030ade8c321feb39 --- /dev/null +++ b/ro/ro_RO/mihai/medium/ro_RO-mihai-medium.onnx.json @@ -0,0 +1,493 @@ +{ + "piper_version": "1.0.0", + "audio": { + "sample_rate": 22050, + "quality": "medium" + }, + "espeak": { + "voice": "ro" + }, + "inference": { + "noise_scale": 0.667, + "length_scale": 1, + "noise_w": 0.8 + }, + "phoneme_type": "espeak", + "phoneme_map": {}, + "phoneme_id_map": { + "_": [ + 0 + ], + "^": [ + 1 + ], + "$": [ + 2 + ], + " ": [ + 3 + ], + "!": [ + 4 + ], + "'": [ + 5 + ], + "(": [ + 6 + ], + ")": [ + 7 + ], + ",": [ + 8 + ], + "-": [ + 9 + ], + ".": [ + 10 + ], + ":": [ + 11 + ], + ";": [ + 12 + ], + "?": [ + 13 + ], + "a": [ + 14 + ], + "b": [ + 15 + ], + "c": [ + 16 + ], + "d": [ + 17 + ], + "e": [ + 18 + ], + "f": [ + 19 + ], + "h": [ + 20 + ], + "i": [ + 21 + ], + "j": [ + 22 + ], + "k": [ + 23 + ], + "l": [ + 24 + ], + "m": [ + 25 + ], + "n": [ + 26 + ], + "o": [ + 27 + ], + "p": [ + 28 + ], + "q": [ + 29 + ], + "r": [ + 30 + ], + "s": [ + 31 + ], + "t": [ + 32 + ], + "u": [ + 33 + ], + "v": [ + 34 + ], + "w": [ + 35 + ], + "x": [ + 36 + ], + "y": [ + 37 + ], + "z": [ + 38 + ], + "æ": [ + 39 + ], + "ç": [ + 40 + ], + "ð": [ + 41 + ], + "ø": [ + 42 + ], + "ħ": [ + 43 + ], + "ŋ": [ + 44 + ], + "œ": [ + 45 + ], + "ǀ": [ + 46 + ], + "ǁ": [ + 47 + ], + "ǂ": [ + 48 + ], + "ǃ": [ + 49 + ], + "ɐ": [ + 50 + ], + "ɑ": [ + 51 + ], + "ɒ": [ + 52 + ], + "ɓ": [ + 53 + ], + "ɔ": [ + 54 + ], + "ɕ": [ + 55 + ], + "ɖ": [ + 56 + ], + "ɗ": [ + 57 + ], + "ɘ": [ + 58 + ], + "ə": [ + 59 + ], + "ɚ": [ + 60 + ], + "ɛ": [ + 61 + ], + "ɜ": [ + 62 + ], + "ɞ": [ + 63 + ], + "ɟ": [ + 64 + ], + "ɠ": [ + 65 + ], + "ɡ": [ + 66 + ], + "ɢ": [ + 67 + ], + "ɣ": [ + 68 + ], + "ɤ": [ + 69 + ], + "ɥ": [ + 70 + ], + "ɦ": [ + 71 + ], + "ɧ": [ + 72 + ], + "ɨ": [ + 73 + ], + "ɪ": [ + 74 + ], + "ɫ": [ + 75 + ], + "ɬ": [ + 76 + ], + "ɭ": [ + 77 + ], + "ɮ": [ + 78 + ], + "ɯ": [ + 79 + ], + "ɰ": [ + 80 + ], + "ɱ": [ + 81 + ], + "ɲ": [ + 82 + ], + "ɳ": [ + 83 + ], + "ɴ": [ + 84 + ], + "ɵ": [ + 85 + ], + "ɶ": [ + 86 + ], + "ɸ": [ + 87 + ], + "ɹ": [ + 88 + ], + "ɺ": [ + 89 + ], + "ɻ": [ + 90 + ], + "ɽ": [ + 91 + ], + "ɾ": [ + 92 + ], + "ʀ": [ + 93 + ], + "ʁ": [ + 94 + ], + "ʂ": [ + 95 + ], + "ʃ": [ + 96 + ], + "ʄ": [ + 97 + ], + "ʈ": [ + 98 + ], + "ʉ": [ + 99 + ], + "ʊ": [ + 100 + ], + "ʋ": [ + 101 + ], + "ʌ": [ + 102 + ], + "ʍ": [ + 103 + ], + "ʎ": [ + 104 + ], + "ʏ": [ + 105 + ], + "ʐ": [ + 106 + ], + "ʑ": [ + 107 + ], + "ʒ": [ + 108 + ], + "ʔ": [ + 109 + ], + "ʕ": [ + 110 + ], + "ʘ": [ + 111 + ], + "ʙ": [ + 112 + ], + "ʛ": [ + 113 + ], + "ʜ": [ + 114 + ], + "ʝ": [ + 115 + ], + "ʟ": [ + 116 + ], + "ʡ": [ + 117 + ], + "ʢ": [ + 118 + ], + "ʲ": [ + 119 + ], + "ˈ": [ + 120 + ], + "ˌ": [ + 121 + ], + "ː": [ + 122 + ], + "ˑ": [ + 123 + ], + "˞": [ + 124 + ], + "β": [ + 125 + ], + "θ": [ + 126 + ], + "χ": [ + 127 + ], + "ᵻ": [ + 128 + ], + "ⱱ": [ + 129 + ], + "0": [ + 130 + ], + "1": [ + 131 + ], + "2": [ + 132 + ], + "3": [ + 133 + ], + "4": [ + 134 + ], + "5": [ + 135 + ], + "6": [ + 136 + ], + "7": [ + 137 + ], + "8": [ + 138 + ], + "9": [ + 139 + ], + "̧": [ + 140 + ], + "̃": [ + 141 + ], + "̪": [ + 142 + ], + "̯": [ + 143 + ], + "̩": [ + 144 + ], + "ʰ": [ + 145 + ], + "ˤ": [ + 146 + ], + "ε": [ + 147 + ], + "↓": [ + 148 + ], + "#": [ + 149 + ], + "\"": [ + 150 + ], + "↑": [ + 151 + ], + "̺": [ + 152 + ], + "̻": [ + 153 + ] + }, + "num_symbols": 256, + "num_speakers": 1, + "speaker_id_map": {}, + "language": { + "code": "ro_RO", + "family": "ro", + "region": "RO", + "name_native": "Română", + "name_english": "Romanian", + "country_english": "Romania" + }, + "dataset": "mihai" +} \ No newline at end of file diff --git a/ro/ro_RO/mihai/medium/samples/speaker_0.mp3 b/ro/ro_RO/mihai/medium/samples/speaker_0.mp3 new file mode 100644 index 0000000000000000000000000000000000000000..8558fa0dd7bdcd7409e020b1a1f57ad7f9c9e70c Binary files /dev/null and b/ro/ro_RO/mihai/medium/samples/speaker_0.mp3 differ