Spaces:
Build error
Build error
Commit
•
1a1be03
0
Parent(s):
Duplicate from internetsignal/Bark-with-Voice-Cloning
Browse files- .gitattributes +40 -0
- Bark.wav +0 -0
- README.md +14 -0
- SE_checkpoint.pth.tar +3 -0
- app.py +313 -0
- best_model.pth.tar +3 -0
- config.json +373 -0
- config_se.json +119 -0
- language_ids.json +5 -0
- ref.wav +3 -0
- requirements.txt +5 -0
- speakers.json +0 -0
.gitattributes
ADDED
@@ -0,0 +1,40 @@
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
1 |
+
*.7z filter=lfs diff=lfs merge=lfs -text
|
2 |
+
*.arrow filter=lfs diff=lfs merge=lfs -text
|
3 |
+
*.bin filter=lfs diff=lfs merge=lfs -text
|
4 |
+
*.bz2 filter=lfs diff=lfs merge=lfs -text
|
5 |
+
*.ckpt filter=lfs diff=lfs merge=lfs -text
|
6 |
+
*.ftz filter=lfs diff=lfs merge=lfs -text
|
7 |
+
*.gz filter=lfs diff=lfs merge=lfs -text
|
8 |
+
*.h5 filter=lfs diff=lfs merge=lfs -text
|
9 |
+
*.joblib filter=lfs diff=lfs merge=lfs -text
|
10 |
+
*.lfs.* filter=lfs diff=lfs merge=lfs -text
|
11 |
+
*.mlmodel filter=lfs diff=lfs merge=lfs -text
|
12 |
+
*.model filter=lfs diff=lfs merge=lfs -text
|
13 |
+
*.msgpack filter=lfs diff=lfs merge=lfs -text
|
14 |
+
*.npy filter=lfs diff=lfs merge=lfs -text
|
15 |
+
*.npz filter=lfs diff=lfs merge=lfs -text
|
16 |
+
*.onnx filter=lfs diff=lfs merge=lfs -text
|
17 |
+
*.ot filter=lfs diff=lfs merge=lfs -text
|
18 |
+
*.parquet filter=lfs diff=lfs merge=lfs -text
|
19 |
+
*.pb filter=lfs diff=lfs merge=lfs -text
|
20 |
+
*.pickle filter=lfs diff=lfs merge=lfs -text
|
21 |
+
*.pkl filter=lfs diff=lfs merge=lfs -text
|
22 |
+
*.pt filter=lfs diff=lfs merge=lfs -text
|
23 |
+
*.pth filter=lfs diff=lfs merge=lfs -text
|
24 |
+
*.rar filter=lfs diff=lfs merge=lfs -text
|
25 |
+
*.safetensors filter=lfs diff=lfs merge=lfs -text
|
26 |
+
saved_model/**/* filter=lfs diff=lfs merge=lfs -text
|
27 |
+
*.tar.* filter=lfs diff=lfs merge=lfs -text
|
28 |
+
*.tflite filter=lfs diff=lfs merge=lfs -text
|
29 |
+
*.tgz filter=lfs diff=lfs merge=lfs -text
|
30 |
+
*.wasm filter=lfs diff=lfs merge=lfs -text
|
31 |
+
*.xz filter=lfs diff=lfs merge=lfs -text
|
32 |
+
*.zip filter=lfs diff=lfs merge=lfs -text
|
33 |
+
*.zst filter=lfs diff=lfs merge=lfs -text
|
34 |
+
*tfevents* filter=lfs diff=lfs merge=lfs -text
|
35 |
+
SE_checkpoint.pth.tar filter=lfs diff=lfs merge=lfs -text
|
36 |
+
best_model.pth.tar filter=lfs diff=lfs merge=lfs -text
|
37 |
+
nana_longest_vocal.wav filter=lfs diff=lfs merge=lfs -text
|
38 |
+
test.wav filter=lfs diff=lfs merge=lfs -text
|
39 |
+
reference.wav filter=lfs diff=lfs merge=lfs -text
|
40 |
+
ref.wav filter=lfs diff=lfs merge=lfs -text
|
Bark.wav
ADDED
Binary file (449 kB). View file
|
|
README.md
ADDED
@@ -0,0 +1,14 @@
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
1 |
+
---
|
2 |
+
title: Bark with Voice Cloning
|
3 |
+
emoji: 📊
|
4 |
+
colorFrom: purple
|
5 |
+
colorTo: purple
|
6 |
+
sdk: gradio
|
7 |
+
sdk_version: 3.27.0
|
8 |
+
app_file: app.py
|
9 |
+
pinned: false
|
10 |
+
license: mit
|
11 |
+
duplicated_from: internetsignal/Bark-with-Voice-Cloning
|
12 |
+
---
|
13 |
+
|
14 |
+
Check out the configuration reference at https://huggingface.co/docs/hub/spaces-config-reference
|
SE_checkpoint.pth.tar
ADDED
@@ -0,0 +1,3 @@
|
|
|
|
|
|
|
|
|
1 |
+
version https://git-lfs.github.com/spec/v1
|
2 |
+
oid sha256:8f96efb20cbeeefd81fd8336d7f0155bf8902f82f9474e58ccb19d9e12345172
|
3 |
+
size 44610930
|
app.py
ADDED
@@ -0,0 +1,313 @@
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
1 |
+
import os
|
2 |
+
|
3 |
+
#os.system("pip install git+https://github.com/suno-ai/bark.git")
|
4 |
+
|
5 |
+
from bark.generation import SUPPORTED_LANGS
|
6 |
+
from bark import SAMPLE_RATE, generate_audio
|
7 |
+
from scipy.io.wavfile import write as write_wav
|
8 |
+
from datetime import datetime
|
9 |
+
|
10 |
+
import shutil
|
11 |
+
import gradio as gr
|
12 |
+
|
13 |
+
import sys
|
14 |
+
|
15 |
+
import string
|
16 |
+
import time
|
17 |
+
import argparse
|
18 |
+
import json
|
19 |
+
|
20 |
+
import numpy as np
|
21 |
+
# import IPython
|
22 |
+
# from IPython.display import Audio
|
23 |
+
|
24 |
+
import torch
|
25 |
+
|
26 |
+
from TTS.tts.utils.synthesis import synthesis
|
27 |
+
from TTS.tts.utils.text.symbols import make_symbols, phonemes, symbols
|
28 |
+
try:
|
29 |
+
from TTS.utils.audio import AudioProcessor
|
30 |
+
except:
|
31 |
+
from TTS.utils.audio import AudioProcessor
|
32 |
+
|
33 |
+
|
34 |
+
from TTS.tts.models import setup_model
|
35 |
+
from TTS.config import load_config
|
36 |
+
from TTS.tts.models.vits import *
|
37 |
+
|
38 |
+
from TTS.tts.utils.speakers import SpeakerManager
|
39 |
+
from pydub import AudioSegment
|
40 |
+
|
41 |
+
# from google.colab import files
|
42 |
+
import librosa
|
43 |
+
|
44 |
+
from scipy.io.wavfile import write, read
|
45 |
+
|
46 |
+
import subprocess
|
47 |
+
|
48 |
+
'''
|
49 |
+
from google.colab import drive
|
50 |
+
drive.mount('/content/drive')
|
51 |
+
src_path = os.path.join(os.path.join(os.path.join(os.path.join(os.getcwd(), 'drive'), 'MyDrive'), 'Colab Notebooks'), 'best_model_latest.pth.tar')
|
52 |
+
dst_path = os.path.join(os.getcwd(), 'best_model.pth.tar')
|
53 |
+
shutil.copy(src_path, dst_path)
|
54 |
+
'''
|
55 |
+
|
56 |
+
TTS_PATH = "TTS/"
|
57 |
+
|
58 |
+
# add libraries into environment
|
59 |
+
sys.path.append(TTS_PATH) # set this if TTS is not installed globally
|
60 |
+
|
61 |
+
# Paths definition
|
62 |
+
|
63 |
+
OUT_PATH = 'out/'
|
64 |
+
|
65 |
+
# create output path
|
66 |
+
os.makedirs(OUT_PATH, exist_ok=True)
|
67 |
+
|
68 |
+
# model vars
|
69 |
+
MODEL_PATH = 'best_model.pth.tar'
|
70 |
+
CONFIG_PATH = 'config.json'
|
71 |
+
TTS_LANGUAGES = "language_ids.json"
|
72 |
+
TTS_SPEAKERS = "speakers.json"
|
73 |
+
USE_CUDA = torch.cuda.is_available()
|
74 |
+
|
75 |
+
# load the config
|
76 |
+
C = load_config(CONFIG_PATH)
|
77 |
+
|
78 |
+
# load the audio processor
|
79 |
+
ap = AudioProcessor(**C.audio)
|
80 |
+
|
81 |
+
speaker_embedding = None
|
82 |
+
|
83 |
+
C.model_args['d_vector_file'] = TTS_SPEAKERS
|
84 |
+
C.model_args['use_speaker_encoder_as_loss'] = False
|
85 |
+
|
86 |
+
model = setup_model(C)
|
87 |
+
model.language_manager.set_language_ids_from_file(TTS_LANGUAGES)
|
88 |
+
# print(model.language_manager.num_languages, model.embedded_language_dim)
|
89 |
+
# print(model.emb_l)
|
90 |
+
cp = torch.load(MODEL_PATH, map_location=torch.device('cpu'))
|
91 |
+
# remove speaker encoder
|
92 |
+
model_weights = cp['model'].copy()
|
93 |
+
for key in list(model_weights.keys()):
|
94 |
+
if "speaker_encoder" in key:
|
95 |
+
del model_weights[key]
|
96 |
+
|
97 |
+
model.load_state_dict(model_weights)
|
98 |
+
|
99 |
+
model.eval()
|
100 |
+
|
101 |
+
if USE_CUDA:
|
102 |
+
model = model.cuda()
|
103 |
+
|
104 |
+
# synthesize voice
|
105 |
+
use_griffin_lim = False
|
106 |
+
|
107 |
+
# Paths definition
|
108 |
+
|
109 |
+
CONFIG_SE_PATH = "config_se.json"
|
110 |
+
CHECKPOINT_SE_PATH = "SE_checkpoint.pth.tar"
|
111 |
+
|
112 |
+
# Load the Speaker encoder
|
113 |
+
|
114 |
+
SE_speaker_manager = SpeakerManager(encoder_model_path=CHECKPOINT_SE_PATH, encoder_config_path=CONFIG_SE_PATH, use_cuda=USE_CUDA)
|
115 |
+
|
116 |
+
# Define helper function
|
117 |
+
|
118 |
+
def compute_spec(ref_file):
|
119 |
+
y, sr = librosa.load(ref_file, sr=ap.sample_rate)
|
120 |
+
spec = ap.spectrogram(y)
|
121 |
+
spec = torch.FloatTensor(spec).unsqueeze(0)
|
122 |
+
return spec
|
123 |
+
|
124 |
+
|
125 |
+
def voice_conversion(ta, ra, da):
|
126 |
+
|
127 |
+
target_audio = 'target.wav'
|
128 |
+
reference_audio = 'reference.wav'
|
129 |
+
driving_audio = 'driving.wav'
|
130 |
+
|
131 |
+
write(target_audio, ta[0], ta[1])
|
132 |
+
write(reference_audio, ra[0], ra[1])
|
133 |
+
write(driving_audio, da[0], da[1])
|
134 |
+
|
135 |
+
# !ffmpeg-normalize $target_audio -nt rms -t=-27 -o $target_audio -ar 16000 -f
|
136 |
+
# !ffmpeg-normalize $reference_audio -nt rms -t=-27 -o $reference_audio -ar 16000 -f
|
137 |
+
# !ffmpeg-normalize $driving_audio -nt rms -t=-27 -o $driving_audio -ar 16000 -f
|
138 |
+
|
139 |
+
files = [target_audio, reference_audio, driving_audio]
|
140 |
+
|
141 |
+
for file in files:
|
142 |
+
subprocess.run(["ffmpeg-normalize", file, "-nt", "rms", "-t=-27", "-o", file, "-ar", "16000", "-f"])
|
143 |
+
|
144 |
+
# ta_ = read(target_audio)
|
145 |
+
|
146 |
+
target_emb = SE_speaker_manager.compute_d_vector_from_clip([target_audio])
|
147 |
+
target_emb = torch.FloatTensor(target_emb).unsqueeze(0)
|
148 |
+
|
149 |
+
driving_emb = SE_speaker_manager.compute_d_vector_from_clip([reference_audio])
|
150 |
+
driving_emb = torch.FloatTensor(driving_emb).unsqueeze(0)
|
151 |
+
|
152 |
+
# Convert the voice
|
153 |
+
|
154 |
+
driving_spec = compute_spec(driving_audio)
|
155 |
+
y_lengths = torch.tensor([driving_spec.size(-1)])
|
156 |
+
if USE_CUDA:
|
157 |
+
ref_wav_voc, _, _ = model.voice_conversion(driving_spec.cuda(), y_lengths.cuda(), driving_emb.cuda(), target_emb.cuda())
|
158 |
+
ref_wav_voc = ref_wav_voc.squeeze().cpu().detach().numpy()
|
159 |
+
else:
|
160 |
+
ref_wav_voc, _, _ = model.voice_conversion(driving_spec, y_lengths, driving_emb, target_emb)
|
161 |
+
ref_wav_voc = ref_wav_voc.squeeze().detach().numpy()
|
162 |
+
|
163 |
+
# print("Reference Audio after decoder:")
|
164 |
+
# IPython.display.display(Audio(ref_wav_voc, rate=ap.sample_rate))
|
165 |
+
|
166 |
+
return (ap.sample_rate, ref_wav_voc)
|
167 |
+
|
168 |
+
def generate_text_to_speech(text_prompt, selected_speaker, text_temp, waveform_temp):
|
169 |
+
audio_array = generate_audio(text_prompt, selected_speaker, text_temp, waveform_temp)
|
170 |
+
|
171 |
+
now = datetime.now()
|
172 |
+
date_str = now.strftime("%m-%d-%Y")
|
173 |
+
time_str = now.strftime("%H-%M-%S")
|
174 |
+
|
175 |
+
outputs_folder = os.path.join(os.getcwd(), "outputs")
|
176 |
+
if not os.path.exists(outputs_folder):
|
177 |
+
os.makedirs(outputs_folder)
|
178 |
+
|
179 |
+
sub_folder = os.path.join(outputs_folder, date_str)
|
180 |
+
if not os.path.exists(sub_folder):
|
181 |
+
os.makedirs(sub_folder)
|
182 |
+
|
183 |
+
file_name = f"audio_{time_str}.wav"
|
184 |
+
file_path = os.path.join(sub_folder, file_name)
|
185 |
+
write_wav(file_path, SAMPLE_RATE, audio_array)
|
186 |
+
|
187 |
+
return file_path
|
188 |
+
|
189 |
+
|
190 |
+
speakers_list = []
|
191 |
+
|
192 |
+
for lang, code in SUPPORTED_LANGS:
|
193 |
+
for n in range(10):
|
194 |
+
speakers_list.append(f"{code}_speaker_{n}")
|
195 |
+
|
196 |
+
examples1 = [["ref.wav", "Bark.wav", "Bark.wav"]]
|
197 |
+
|
198 |
+
with gr.Blocks() as demo:
|
199 |
+
gr.Markdown(
|
200 |
+
f""" # <center>🐶🎶🥳 - Bark with Voice Cloning</center>
|
201 |
+
|
202 |
+
### <center>🤗 - Powered by [Bark](https://huggingface.co/spaces/suno/bark) and [YourTTS](https://github.com/Edresson/YourTTS). Inspired by [bark-webui](https://github.com/makawy7/bark-webui).</center>
|
203 |
+
1. You can duplicate and use it with a GPU: <a href="https://huggingface.co/spaces/{os.getenv('SPACE_ID')}?duplicate=true"><img style="display: inline; margin-top: 0em; margin-bottom: 0em" src="https://bit.ly/3gLdBN6" alt="Duplicate Space" /></a>
|
204 |
+
2. First use Bark to generate audio from text and then use YourTTS to get new audio in a custom voice you like. Easy to use!
|
205 |
+
3. For voice cloning, longer reference audio (~90s) will generally lead to better quality of the cloned speech. Also, please make sure the input audio generated by Bark is not too short.
|
206 |
+
"""
|
207 |
+
)
|
208 |
+
|
209 |
+
with gr.Row().style(equal_height=True):
|
210 |
+
inp1 = gr.Textbox(label="Input Text", lines=4, placeholder="Enter text here...")
|
211 |
+
|
212 |
+
inp3 = gr.Slider(
|
213 |
+
0.1,
|
214 |
+
1.0,
|
215 |
+
value=0.7,
|
216 |
+
label="Generation Temperature",
|
217 |
+
info="1.0 more diverse, 0.1 more conservative",
|
218 |
+
)
|
219 |
+
|
220 |
+
inp4 = gr.Slider(
|
221 |
+
0.1, 1.0, value=0.7, label="Waveform Temperature", info="1.0 more diverse, 0.1 more conservative"
|
222 |
+
)
|
223 |
+
with gr.Row().style(equal_height=True):
|
224 |
+
|
225 |
+
inp2 = gr.Dropdown(speakers_list, value=speakers_list[1], label="Acoustic Prompt")
|
226 |
+
|
227 |
+
button = gr.Button("Generate using Bark")
|
228 |
+
|
229 |
+
out1 = gr.Audio(label="Generated Audio")
|
230 |
+
|
231 |
+
button.click(generate_text_to_speech, [inp1, inp2, inp3, inp4], [out1])
|
232 |
+
|
233 |
+
|
234 |
+
with gr.Row().style(equal_height=True):
|
235 |
+
inp5 = gr.Audio(label="Upload Reference Audio for Voice Cloning Here")
|
236 |
+
inp6 = out1
|
237 |
+
inp7 = out1
|
238 |
+
|
239 |
+
btn = gr.Button("Generate using YourTTS")
|
240 |
+
out2 = gr.Audio(label="Generated Audio in a Custom Voice")
|
241 |
+
|
242 |
+
btn.click(voice_conversion, [inp5, inp6, inp7], [out2])
|
243 |
+
|
244 |
+
gr.Examples(examples=examples1, fn=voice_conversion, inputs=[inp5, inp6, inp7],
|
245 |
+
outputs=[out2], cache_examples=True)
|
246 |
+
|
247 |
+
gr.Markdown(
|
248 |
+
""" ### <center>NOTE: Please do not generate any audio that is potentially harmful to any person or organization❗</center>
|
249 |
+
|
250 |
+
"""
|
251 |
+
)
|
252 |
+
gr.Markdown(
|
253 |
+
"""
|
254 |
+
### <center>😄 - You may also apply [VoiceFixer](https://huggingface.co/spaces/Kevin676/VoiceFixer) to the generated audio in order to enhance the speech.</center>
|
255 |
+
## 🌎 Foreign Language
|
256 |
+
Bark supports various languages out-of-the-box and automatically determines language from input text. \
|
257 |
+
When prompted with code-switched text, Bark will even attempt to employ the native accent for the respective languages in the same voice.
|
258 |
+
Try the prompt:
|
259 |
+
```
|
260 |
+
Buenos días Miguel. Tu colega piensa que tu alemán es extremadamente malo. But I suppose your english isn't terrible.
|
261 |
+
```
|
262 |
+
## 🤭 Non-Speech Sounds
|
263 |
+
Below is a list of some known non-speech sounds, but we are finding more every day. \
|
264 |
+
Please let us know if you find patterns that work particularly well on Discord!
|
265 |
+
* [laughter]
|
266 |
+
* [laughs]
|
267 |
+
* [sighs]
|
268 |
+
* [music]
|
269 |
+
* [gasps]
|
270 |
+
* [clears throat]
|
271 |
+
* — or ... for hesitations
|
272 |
+
* ♪ for song lyrics
|
273 |
+
* capitalization for emphasis of a word
|
274 |
+
* MAN/WOMAN: for bias towards speaker
|
275 |
+
Try the prompt:
|
276 |
+
```
|
277 |
+
" [clears throat] Hello, my name is Suno. And, uh — and I like pizza. [laughs] But I also have other interests such as... ♪ singing ♪."
|
278 |
+
```
|
279 |
+
## 🎶 Music
|
280 |
+
Bark can generate all types of audio, and, in principle, doesn't see a difference between speech and music. \
|
281 |
+
Sometimes Bark chooses to generate text as music, but you can help it out by adding music notes around your lyrics.
|
282 |
+
Try the prompt:
|
283 |
+
```
|
284 |
+
♪ In the jungle, the mighty jungle, the lion barks tonight ♪
|
285 |
+
```
|
286 |
+
## 🧬 Voice Cloning
|
287 |
+
Bark has the capability to fully clone voices - including tone, pitch, emotion and prosody. \
|
288 |
+
The model also attempts to preserve music, ambient noise, etc. from input audio. \
|
289 |
+
However, to mitigate misuse of this technology, we limit the audio history prompts to a limited set of Suno-provided, fully synthetic options to choose from.
|
290 |
+
## 👥 Speaker Prompts
|
291 |
+
You can provide certain speaker prompts such as NARRATOR, MAN, WOMAN, etc. \
|
292 |
+
Please note that these are not always respected, especially if a conflicting audio history prompt is given.
|
293 |
+
Try the prompt:
|
294 |
+
```
|
295 |
+
WOMAN: I would like an oatmilk latte please.
|
296 |
+
MAN: Wow, that's expensive!
|
297 |
+
```
|
298 |
+
## Details
|
299 |
+
Bark model by [Suno](https://suno.ai/), including official [code](https://github.com/suno-ai/bark) and model weights. \
|
300 |
+
Gradio demo supported by 🤗 Hugging Face. Bark is licensed under a non-commercial license: CC-BY 4.0 NC, see details on [GitHub](https://github.com/suno-ai/bark).
|
301 |
+
|
302 |
+
"""
|
303 |
+
)
|
304 |
+
|
305 |
+
|
306 |
+
gr.HTML('''
|
307 |
+
<div class="footer">
|
308 |
+
<p>🎶🖼️🎡 - It’s the intersection of technology and liberal arts that makes our hearts sing — Steve Jobs
|
309 |
+
</p>
|
310 |
+
</div>
|
311 |
+
''')
|
312 |
+
|
313 |
+
demo.queue().launch(show_error=True)
|
best_model.pth.tar
ADDED
@@ -0,0 +1,3 @@
|
|
|
|
|
|
|
|
|
1 |
+
version https://git-lfs.github.com/spec/v1
|
2 |
+
oid sha256:017bfd8907c80bb5857d65d0223f0e4e4b9d699ef52e2a853d9cc7eb7e308cf0
|
3 |
+
size 379957289
|
config.json
ADDED
@@ -0,0 +1,373 @@
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
1 |
+
{
|
2 |
+
"model": "vits",
|
3 |
+
"run_name": "vits_tts-portuguese",
|
4 |
+
"run_description": "",
|
5 |
+
"epochs": 1000,
|
6 |
+
"batch_size": 52,
|
7 |
+
"eval_batch_size": 52,
|
8 |
+
"mixed_precision": false,
|
9 |
+
"scheduler_after_epoch": true,
|
10 |
+
"run_eval": true,
|
11 |
+
"test_delay_epochs": -1,
|
12 |
+
"print_eval": true,
|
13 |
+
"dashboard_logger": "tensorboard",
|
14 |
+
"print_step": 25,
|
15 |
+
"plot_step": 100,
|
16 |
+
"model_param_stats": false,
|
17 |
+
"project_name": null,
|
18 |
+
"log_model_step": 10000,
|
19 |
+
"wandb_entity": null,
|
20 |
+
"save_step": 10000,
|
21 |
+
"checkpoint": true,
|
22 |
+
"keep_all_best": false,
|
23 |
+
"keep_after": 10000,
|
24 |
+
"num_loader_workers": 4,
|
25 |
+
"num_eval_loader_workers": 4,
|
26 |
+
"use_noise_augment": false,
|
27 |
+
"use_language_weighted_sampler": true,
|
28 |
+
"output_path": "../checkpoints/VITS-multilingual/VITS_fixes/new/new-SE/use_noise_aument_false/xlarge-ZS-PT-VCTK/pt-en+LibriTTS-fr/speaker_encoder_as_loss_9_alpha/mixed-p-false-bug-SDP-fixed/",
|
29 |
+
"distributed_backend": "nccl",
|
30 |
+
"distributed_url": "tcp://localhost:54321",
|
31 |
+
"audio": {
|
32 |
+
"fft_size": 1024,
|
33 |
+
"win_length": 1024,
|
34 |
+
"hop_length": 256,
|
35 |
+
"frame_shift_ms": null,
|
36 |
+
"frame_length_ms": null,
|
37 |
+
"stft_pad_mode": "reflect",
|
38 |
+
"sample_rate": 16000,
|
39 |
+
"resample": false,
|
40 |
+
"preemphasis": 0.0,
|
41 |
+
"ref_level_db": 20,
|
42 |
+
"do_sound_norm": false,
|
43 |
+
"log_func": "np.log",
|
44 |
+
"do_trim_silence": true,
|
45 |
+
"trim_db": 45,
|
46 |
+
"power": 1.5,
|
47 |
+
"griffin_lim_iters": 60,
|
48 |
+
"num_mels": 80,
|
49 |
+
"mel_fmin": 0.0,
|
50 |
+
"mel_fmax": null,
|
51 |
+
"spec_gain": 1,
|
52 |
+
"do_amp_to_db_linear": false,
|
53 |
+
"do_amp_to_db_mel": true,
|
54 |
+
"signal_norm": false,
|
55 |
+
"min_level_db": -100,
|
56 |
+
"symmetric_norm": true,
|
57 |
+
"max_norm": 4.0,
|
58 |
+
"clip_norm": true,
|
59 |
+
"stats_path": null
|
60 |
+
},
|
61 |
+
"use_phonemes": false,
|
62 |
+
"use_espeak_phonemes": false,
|
63 |
+
"phoneme_language": "pt-br",
|
64 |
+
"compute_input_seq_cache": false,
|
65 |
+
"text_cleaner": "multilingual_cleaners",
|
66 |
+
"enable_eos_bos_chars": false,
|
67 |
+
"test_sentences_file": "",
|
68 |
+
"phoneme_cache_path": null,
|
69 |
+
"characters": {
|
70 |
+
"pad": "_",
|
71 |
+
"eos": "&",
|
72 |
+
"bos": "*",
|
73 |
+
"characters": "ABCDEFGHIJKLMNOPQRSTUVWXYZabcdefghijklmnopqrstuvwxyz\u00af\u00b7\u00df\u00e0\u00e1\u00e2\u00e3\u00e4\u00e6\u00e7\u00e8\u00e9\u00ea\u00eb\u00ec\u00ed\u00ee\u00ef\u00f1\u00f2\u00f3\u00f4\u00f5\u00f6\u00f9\u00fa\u00fb\u00fc\u00ff\u0101\u0105\u0107\u0113\u0119\u011b\u012b\u0131\u0142\u0144\u014d\u0151\u0153\u015b\u016b\u0171\u017a\u017c\u01ce\u01d0\u01d2\u01d4\u0430\u0431\u0432\u0433\u0434\u0435\u0436\u0437\u0438\u0439\u043a\u043b\u043c\u043d\u043e\u043f\u0440\u0441\u0442\u0443\u0444\u0445\u0446\u0447\u0448\u0449\u044a\u044b\u044c\u044d\u044e\u044f\u0451\u0454\u0456\u0457\u0491\u2013!'(),-.:;? ",
|
74 |
+
"punctuations": "!'(),-.:;? ",
|
75 |
+
"phonemes": "iy\u0268\u0289\u026fu\u026a\u028f\u028ae\u00f8\u0258\u0259\u0275\u0264o\u025b\u0153\u025c\u025e\u028c\u0254\u00e6\u0250a\u0276\u0251\u0252\u1d7b\u0298\u0253\u01c0\u0257\u01c3\u0284\u01c2\u0260\u01c1\u029bpbtd\u0288\u0256c\u025fk\u0261q\u0262\u0294\u0274\u014b\u0272\u0273n\u0271m\u0299r\u0280\u2c71\u027e\u027d\u0278\u03b2fv\u03b8\u00f0sz\u0283\u0292\u0282\u0290\u00e7\u029dx\u0263\u03c7\u0281\u0127\u0295h\u0266\u026c\u026e\u028b\u0279\u027bj\u0270l\u026d\u028e\u029f\u02c8\u02cc\u02d0\u02d1\u028dw\u0265\u029c\u02a2\u02a1\u0255\u0291\u027a\u0267\u025a\u02de\u026b'\u0303' ",
|
76 |
+
"unique": true
|
77 |
+
},
|
78 |
+
"batch_group_size": 0,
|
79 |
+
"loss_masking": null,
|
80 |
+
"min_seq_len": 90,
|
81 |
+
"max_seq_len": 270,
|
82 |
+
"compute_f0": false,
|
83 |
+
"compute_linear_spec": true,
|
84 |
+
"add_blank": true,
|
85 |
+
"datasets": [
|
86 |
+
{
|
87 |
+
"name": "vctk",
|
88 |
+
"path": "../../datasets/VCTK-Corpus-removed-silence_16Khz/",
|
89 |
+
"meta_file_train": null,
|
90 |
+
"ununsed_speakers": [
|
91 |
+
"p225",
|
92 |
+
"p234",
|
93 |
+
"p238",
|
94 |
+
"p245",
|
95 |
+
"p248",
|
96 |
+
"p261",
|
97 |
+
"p294",
|
98 |
+
"p302",
|
99 |
+
"p326",
|
100 |
+
"p335",
|
101 |
+
"p347"
|
102 |
+
],
|
103 |
+
"language": "en",
|
104 |
+
"meta_file_val": null,
|
105 |
+
"meta_file_attn_mask": ""
|
106 |
+
},
|
107 |
+
{
|
108 |
+
"name": "libri_tts",
|
109 |
+
"path": "../../datasets/LibriTTS/LibriTTS/dataset-preprocessed-clean-100-and-360/dataset-22k/",
|
110 |
+
"meta_file_train": "metadata_all.csv",
|
111 |
+
"ununsed_speakers": null,
|
112 |
+
"language": "en",
|
113 |
+
"meta_file_val": "dev-clean_500.csv",
|
114 |
+
"meta_file_attn_mask": ""
|
115 |
+
},
|
116 |
+
{
|
117 |
+
"name": "brspeech",
|
118 |
+
"path": "../../datasets/TTS-Portuguese-Corpus_16khz/",
|
119 |
+
"meta_file_train": "train_TTS-Portuguese_Corpus_metadata.csv",
|
120 |
+
"ununsed_speakers": null,
|
121 |
+
"language": "pt-br",
|
122 |
+
"meta_file_val": "eval_TTS-Portuguese_Corpus_metadata.csv",
|
123 |
+
"meta_file_attn_mask": ""
|
124 |
+
},
|
125 |
+
{
|
126 |
+
"name": "mailabs",
|
127 |
+
"path": "../../datasets/M-AILABS/fr_FR",
|
128 |
+
"meta_file_train": "",
|
129 |
+
"ununsed_speakers": null,
|
130 |
+
"language": "fr-fr",
|
131 |
+
"meta_file_val": null,
|
132 |
+
"meta_file_attn_mask": null
|
133 |
+
}
|
134 |
+
],
|
135 |
+
"optimizer": "AdamW",
|
136 |
+
"optimizer_params": {
|
137 |
+
"betas": [
|
138 |
+
0.8,
|
139 |
+
0.99
|
140 |
+
],
|
141 |
+
"eps": 1e-09,
|
142 |
+
"weight_decay": 0.01
|
143 |
+
},
|
144 |
+
"lr_scheduler": "",
|
145 |
+
"lr_scheduler_params": null,
|
146 |
+
"test_sentences": [
|
147 |
+
[
|
148 |
+
"It took me quite a long time to develop a voice, and now that I have it I'm not going to be silent.",
|
149 |
+
"VCTK_p225",
|
150 |
+
null,
|
151 |
+
"en"
|
152 |
+
],
|
153 |
+
[
|
154 |
+
"It took me quite a long time to develop a voice, and now that I have it I'm not going to be silent.",
|
155 |
+
"ED",
|
156 |
+
null,
|
157 |
+
"en"
|
158 |
+
],
|
159 |
+
[
|
160 |
+
"It took me quite a long time to develop a voice, and now that I have it I'm not going to be silent.",
|
161 |
+
"bernard",
|
162 |
+
null,
|
163 |
+
"en"
|
164 |
+
],
|
165 |
+
[
|
166 |
+
"This cake is great. It's so delicious and moist.",
|
167 |
+
"VCTK_p234",
|
168 |
+
null,
|
169 |
+
"en"
|
170 |
+
],
|
171 |
+
[
|
172 |
+
"This cake is great. It's so delicious and moist.",
|
173 |
+
"ED",
|
174 |
+
null,
|
175 |
+
"en"
|
176 |
+
],
|
177 |
+
[
|
178 |
+
"This cake is great. It's so delicious and moist.",
|
179 |
+
"ezwa",
|
180 |
+
null,
|
181 |
+
"en"
|
182 |
+
],
|
183 |
+
[
|
184 |
+
"Hoje \u00e9 fundamental encontrar a raz\u00e3o da exist\u00eancia humana.",
|
185 |
+
"ED",
|
186 |
+
null,
|
187 |
+
"pt-br"
|
188 |
+
],
|
189 |
+
[
|
190 |
+
"Hoje \u00e9 fundamental encontrar a raz\u00e3o da exist\u00eancia humana.",
|
191 |
+
"VCTK_p238",
|
192 |
+
null,
|
193 |
+
"pt-br"
|
194 |
+
],
|
195 |
+
[
|
196 |
+
"Hoje \u00e9 fundamental encontrar a raz\u00e3o da exist\u00eancia humana.",
|
197 |
+
"gilles_g_le_blanc",
|
198 |
+
null,
|
199 |
+
"pt-br"
|
200 |
+
],
|
201 |
+
[
|
202 |
+
"Em muitas cidades a popula\u00e7\u00e3o est\u00e1 diminuindo.",
|
203 |
+
"ED",
|
204 |
+
null,
|
205 |
+
"pt-br"
|
206 |
+
],
|
207 |
+
[
|
208 |
+
"Em muitas cidades a popula\u00e7\u00e3o est\u00e1 diminuindo.",
|
209 |
+
"VCTK_p245",
|
210 |
+
null,
|
211 |
+
"pt-br"
|
212 |
+
],
|
213 |
+
[
|
214 |
+
"Em muitas cidades a popula\u00e7\u00e3o est\u00e1 diminuindo.",
|
215 |
+
"nadine_eckert_boulet",
|
216 |
+
null,
|
217 |
+
"pt-br"
|
218 |
+
],
|
219 |
+
[
|
220 |
+
"Il m'a fallu beaucoup de temps pour d\u00e9velopper une voix, et maintenant que je l'ai, je ne vais pas me taire.",
|
221 |
+
"VCTK_p245",
|
222 |
+
null,
|
223 |
+
"fr-fr"
|
224 |
+
],
|
225 |
+
[
|
226 |
+
"Il m'a fallu beaucoup de temps pour d\u00e9velopper une voix, et maintenant que je l'ai, je ne vais pas me taire.",
|
227 |
+
"ED",
|
228 |
+
null,
|
229 |
+
"fr-fr"
|
230 |
+
],
|
231 |
+
[
|
232 |
+
"Il m'a fallu beaucoup de temps pour d\u00e9velopper une voix, et maintenant que je l'ai, je ne vais pas me taire.",
|
233 |
+
"ezwa",
|
234 |
+
null,
|
235 |
+
"fr-fr"
|
236 |
+
],
|
237 |
+
[
|
238 |
+
"Il m'a fallu beaucoup de temps pour d\u00e9velopper une voix, et maintenant que je l'ai, je ne vais pas me taire.",
|
239 |
+
"bernard",
|
240 |
+
null,
|
241 |
+
"fr-fr"
|
242 |
+
],
|
243 |
+
[
|
244 |
+
"Il m'a fallu beaucoup de temps pour d\u00e9velopper une voix, et maintenant que je l'ai, je ne vais pas me taire.",
|
245 |
+
"gilles_g_le_blanc",
|
246 |
+
null,
|
247 |
+
"fr-fr"
|
248 |
+
],
|
249 |
+
[
|
250 |
+
"Il m'a fallu beaucoup de temps pour d\u00e9velopper une voix, et maintenant que je l'ai, je ne vais pas me taire.",
|
251 |
+
"nadine_eckert_boulet",
|
252 |
+
null,
|
253 |
+
"fr-fr"
|
254 |
+
],
|
255 |
+
[
|
256 |
+
"Il m'a fallu beaucoup de temps pour d\u00e9velopper une voix, et maintenant que je l'ai, je ne vais pas me taire.",
|
257 |
+
"zeckou",
|
258 |
+
null,
|
259 |
+
"fr-fr"
|
260 |
+
]
|
261 |
+
],
|
262 |
+
"use_speaker_embedding": true,
|
263 |
+
"use_d_vector_file": true,
|
264 |
+
"d_vector_dim": 512,
|
265 |
+
"model_args": {
|
266 |
+
"num_chars": 165,
|
267 |
+
"out_channels": 513,
|
268 |
+
"spec_segment_size": 62,
|
269 |
+
"hidden_channels": 192,
|
270 |
+
"hidden_channels_ffn_text_encoder": 768,
|
271 |
+
"num_heads_text_encoder": 2,
|
272 |
+
"num_layers_text_encoder": 10,
|
273 |
+
"kernel_size_text_encoder": 3,
|
274 |
+
"dropout_p_text_encoder": 0.1,
|
275 |
+
"dropout_p_duration_predictor": 0.5,
|
276 |
+
"kernel_size_posterior_encoder": 5,
|
277 |
+
"dilation_rate_posterior_encoder": 1,
|
278 |
+
"num_layers_posterior_encoder": 16,
|
279 |
+
"kernel_size_flow": 5,
|
280 |
+
"dilation_rate_flow": 1,
|
281 |
+
"num_layers_flow": 4,
|
282 |
+
"resblock_type_decoder": 1,
|
283 |
+
"resblock_kernel_sizes_decoder": [
|
284 |
+
3,
|
285 |
+
7,
|
286 |
+
11
|
287 |
+
],
|
288 |
+
"resblock_dilation_sizes_decoder": [
|
289 |
+
[
|
290 |
+
1,
|
291 |
+
3,
|
292 |
+
5
|
293 |
+
],
|
294 |
+
[
|
295 |
+
1,
|
296 |
+
3,
|
297 |
+
5
|
298 |
+
],
|
299 |
+
[
|
300 |
+
1,
|
301 |
+
3,
|
302 |
+
5
|
303 |
+
]
|
304 |
+
],
|
305 |
+
"upsample_rates_decoder": [
|
306 |
+
8,
|
307 |
+
8,
|
308 |
+
2,
|
309 |
+
2
|
310 |
+
],
|
311 |
+
"upsample_initial_channel_decoder": 512,
|
312 |
+
"upsample_kernel_sizes_decoder": [
|
313 |
+
16,
|
314 |
+
16,
|
315 |
+
4,
|
316 |
+
4
|
317 |
+
],
|
318 |
+
"use_sdp": true,
|
319 |
+
"noise_scale": 1.0,
|
320 |
+
"inference_noise_scale": 0.667,
|
321 |
+
"length_scale": 1,
|
322 |
+
"noise_scale_dp": 1.0,
|
323 |
+
"inference_noise_scale_dp": 0.8,
|
324 |
+
"max_inference_len": null,
|
325 |
+
"init_discriminator": true,
|
326 |
+
"use_spectral_norm_disriminator": false,
|
327 |
+
"use_speaker_embedding": true,
|
328 |
+
"num_speakers": 1244,
|
329 |
+
"speakers_file": null,
|
330 |
+
"d_vector_file": "../speaker_embeddings/new-SE/VCTK-LibriTTS+TTS-PT+MAILABS-FR/speakers.json",
|
331 |
+
"speaker_embedding_channels": 512,
|
332 |
+
"use_d_vector_file": true,
|
333 |
+
"d_vector_dim": 512,
|
334 |
+
"detach_dp_input": true,
|
335 |
+
"use_language_embedding": true,
|
336 |
+
"embedded_language_dim": 4,
|
337 |
+
"num_languages": 3,
|
338 |
+
"use_speaker_encoder_as_loss": true,
|
339 |
+
"speaker_encoder_config_path": "../checkpoints/Speaker_Encoder/Resnet-original-paper/config.json",
|
340 |
+
"speaker_encoder_model_path": "../checkpoints/Speaker_Encoder/Resnet-original-paper/converted_checkpoint.pth.tar",
|
341 |
+
"fine_tuning_mode": 0,
|
342 |
+
"freeze_encoder": false,
|
343 |
+
"freeze_DP": false,
|
344 |
+
"freeze_PE": false,
|
345 |
+
"freeze_flow_decoder": false,
|
346 |
+
"freeze_waveform_decoder": false
|
347 |
+
},
|
348 |
+
"grad_clip": [
|
349 |
+
5.0,
|
350 |
+
5.0
|
351 |
+
],
|
352 |
+
"lr_gen": 0.0002,
|
353 |
+
"lr_disc": 0.0002,
|
354 |
+
"lr_scheduler_gen": "ExponentialLR",
|
355 |
+
"lr_scheduler_gen_params": {
|
356 |
+
"gamma": 0.999875,
|
357 |
+
"last_epoch": -1
|
358 |
+
},
|
359 |
+
"lr_scheduler_disc": "ExponentialLR",
|
360 |
+
"lr_scheduler_disc_params": {
|
361 |
+
"gamma": 0.999875,
|
362 |
+
"last_epoch": -1
|
363 |
+
},
|
364 |
+
"kl_loss_alpha": 1.0,
|
365 |
+
"disc_loss_alpha": 1.0,
|
366 |
+
"gen_loss_alpha": 1.0,
|
367 |
+
"feat_loss_alpha": 1.0,
|
368 |
+
"mel_loss_alpha": 45.0,
|
369 |
+
"dur_loss_alpha": 1.0,
|
370 |
+
"speaker_encoder_loss_alpha": 9.0,
|
371 |
+
"return_wav": true,
|
372 |
+
"r": 1
|
373 |
+
}
|
config_se.json
ADDED
@@ -0,0 +1,119 @@
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
1 |
+
{
|
2 |
+
"model": "speaker_encoder",
|
3 |
+
"run_name": "speaker_encoder",
|
4 |
+
"run_description": "resnet speaker encoder trained with commonvoice all languages dev and train, Voxceleb 1 dev and Voxceleb 2 dev",
|
5 |
+
"epochs": 100000,
|
6 |
+
"batch_size": null,
|
7 |
+
"eval_batch_size": null,
|
8 |
+
"mixed_precision": false,
|
9 |
+
"run_eval": true,
|
10 |
+
"test_delay_epochs": 0,
|
11 |
+
"print_eval": false,
|
12 |
+
"print_step": 50,
|
13 |
+
"tb_plot_step": 100,
|
14 |
+
"tb_model_param_stats": false,
|
15 |
+
"save_step": 1000,
|
16 |
+
"checkpoint": true,
|
17 |
+
"keep_all_best": false,
|
18 |
+
"keep_after": 10000,
|
19 |
+
"num_loader_workers": 8,
|
20 |
+
"num_val_loader_workers": 0,
|
21 |
+
"use_noise_augment": false,
|
22 |
+
"output_path": "../checkpoints/speaker_encoder/language_balanced/normalized/angleproto-4-samples-by-speakers/",
|
23 |
+
"distributed_backend": "nccl",
|
24 |
+
"distributed_url": "tcp://localhost:54321",
|
25 |
+
"audio": {
|
26 |
+
"fft_size": 512,
|
27 |
+
"win_length": 400,
|
28 |
+
"hop_length": 160,
|
29 |
+
"frame_shift_ms": null,
|
30 |
+
"frame_length_ms": null,
|
31 |
+
"stft_pad_mode": "reflect",
|
32 |
+
"sample_rate": 16000,
|
33 |
+
"resample": false,
|
34 |
+
"preemphasis": 0.97,
|
35 |
+
"ref_level_db": 20,
|
36 |
+
"do_sound_norm": false,
|
37 |
+
"do_trim_silence": false,
|
38 |
+
"trim_db": 60,
|
39 |
+
"power": 1.5,
|
40 |
+
"griffin_lim_iters": 60,
|
41 |
+
"num_mels": 64,
|
42 |
+
"mel_fmin": 0.0,
|
43 |
+
"mel_fmax": 8000.0,
|
44 |
+
"spec_gain": 20,
|
45 |
+
"signal_norm": false,
|
46 |
+
"min_level_db": -100,
|
47 |
+
"symmetric_norm": false,
|
48 |
+
"max_norm": 4.0,
|
49 |
+
"clip_norm": false,
|
50 |
+
"stats_path": null
|
51 |
+
},
|
52 |
+
"datasets": [
|
53 |
+
{
|
54 |
+
"name": "voxceleb2",
|
55 |
+
"path": "/workspace/scratch/ecasanova/datasets/VoxCeleb/vox2_dev_aac/",
|
56 |
+
"meta_file_train": null,
|
57 |
+
"ununsed_speakers": null,
|
58 |
+
"meta_file_val": null,
|
59 |
+
"meta_file_attn_mask": "",
|
60 |
+
"language": "voxceleb"
|
61 |
+
}
|
62 |
+
],
|
63 |
+
"model_params": {
|
64 |
+
"model_name": "resnet",
|
65 |
+
"input_dim": 64,
|
66 |
+
"use_torch_spec": true,
|
67 |
+
"log_input": true,
|
68 |
+
"proj_dim": 512
|
69 |
+
},
|
70 |
+
"audio_augmentation": {
|
71 |
+
"p": 0.5,
|
72 |
+
"rir": {
|
73 |
+
"rir_path": "/workspace/store/ecasanova/ComParE/RIRS_NOISES/simulated_rirs/",
|
74 |
+
"conv_mode": "full"
|
75 |
+
},
|
76 |
+
"additive": {
|
77 |
+
"sounds_path": "/workspace/store/ecasanova/ComParE/musan/",
|
78 |
+
"speech": {
|
79 |
+
"min_snr_in_db": 13,
|
80 |
+
"max_snr_in_db": 20,
|
81 |
+
"min_num_noises": 1,
|
82 |
+
"max_num_noises": 1
|
83 |
+
},
|
84 |
+
"noise": {
|
85 |
+
"min_snr_in_db": 0,
|
86 |
+
"max_snr_in_db": 15,
|
87 |
+
"min_num_noises": 1,
|
88 |
+
"max_num_noises": 1
|
89 |
+
},
|
90 |
+
"music": {
|
91 |
+
"min_snr_in_db": 5,
|
92 |
+
"max_snr_in_db": 15,
|
93 |
+
"min_num_noises": 1,
|
94 |
+
"max_num_noises": 1
|
95 |
+
}
|
96 |
+
},
|
97 |
+
"gaussian": {
|
98 |
+
"p": 0.0,
|
99 |
+
"min_amplitude": 0.0,
|
100 |
+
"max_amplitude": 1e-05
|
101 |
+
}
|
102 |
+
},
|
103 |
+
"storage": {
|
104 |
+
"sample_from_storage_p": 0.5,
|
105 |
+
"storage_size": 40
|
106 |
+
},
|
107 |
+
"max_train_step": 1000000,
|
108 |
+
"loss": "angleproto",
|
109 |
+
"grad_clip": 3.0,
|
110 |
+
"lr": 0.0001,
|
111 |
+
"lr_decay": false,
|
112 |
+
"warmup_steps": 4000,
|
113 |
+
"wd": 1e-06,
|
114 |
+
"steps_plot_stats": 100,
|
115 |
+
"num_speakers_in_batch": 100,
|
116 |
+
"num_utters_per_speaker": 4,
|
117 |
+
"skip_speakers": true,
|
118 |
+
"voice_len": 2.0
|
119 |
+
}
|
language_ids.json
ADDED
@@ -0,0 +1,5 @@
|
|
|
|
|
|
|
|
|
|
|
|
|
1 |
+
{
|
2 |
+
"en": 0,
|
3 |
+
"fr-fr": 1,
|
4 |
+
"pt-br": 2
|
5 |
+
}
|
ref.wav
ADDED
@@ -0,0 +1,3 @@
|
|
|
|
|
|
|
|
|
1 |
+
version https://git-lfs.github.com/spec/v1
|
2 |
+
oid sha256:056ae4d5aba4c623ff5815c7425d0431db80be276b0582432aa0aa8ded6f90f2
|
3 |
+
size 15996512
|
requirements.txt
ADDED
@@ -0,0 +1,5 @@
|
|
|
|
|
|
|
|
|
|
|
|
|
1 |
+
git+https://github.com/suno-ai/bark.git
|
2 |
+
git+https://github.com/Edresson/Coqui-TTS@multilingual-torchaudio-SE
|
3 |
+
torchaudio
|
4 |
+
pydub
|
5 |
+
ffmpeg-normalize==1.21.0
|
speakers.json
ADDED
The diff for this file is too large to render.
See raw diff
|
|