Update app.py
Browse files
app.py
CHANGED
@@ -2,19 +2,31 @@ import gradio as gr
|
|
2 |
import numpy as np
|
3 |
import torch
|
4 |
from datasets import load_dataset
|
5 |
-
|
6 |
from transformers import SpeechT5ForTextToSpeech, SpeechT5HifiGan, SpeechT5Processor, pipeline
|
|
|
7 |
|
8 |
|
9 |
device = "cuda:0" if torch.cuda.is_available() else "cpu"
|
10 |
|
11 |
# load speech translation checkpoint
|
12 |
-
asr_pipe = pipeline("automatic-speech-recognition", model="openai/whisper-base", device=device)
|
|
|
|
|
|
|
|
|
|
|
13 |
|
14 |
# load text-to-speech checkpoint and speaker embeddings
|
15 |
-
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
16 |
|
17 |
-
model = SpeechT5ForTextToSpeech.from_pretrained("Matthijs/mms-tts-fra").to(device)
|
18 |
vocoder = SpeechT5HifiGan.from_pretrained("microsoft/speecht5_hifigan").to(device)
|
19 |
|
20 |
embeddings_dataset = load_dataset("Matthijs/cmu-arctic-xvectors", split="validation")
|
@@ -22,23 +34,40 @@ speaker_embeddings = torch.tensor(embeddings_dataset[7306]["xvector"]).unsqueeze
|
|
22 |
|
23 |
|
24 |
def translate(audio):
|
25 |
-
|
26 |
-
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
27 |
|
28 |
|
29 |
def synthesise(text):
|
30 |
-
|
31 |
-
|
32 |
-
speech = model.generate_speech(inputs["input_ids"].to(device), speaker_embeddings.to(device))
|
33 |
-
|
34 |
-
|
35 |
-
|
36 |
-
|
37 |
-
|
|
|
|
|
|
|
|
|
|
|
|
|
38 |
|
39 |
|
40 |
def speech_to_speech_translation(audio):
|
41 |
translated_text = translate(audio)
|
|
|
42 |
synthesised_speech = synthesise(translated_text)
|
43 |
synthesised_speech = (synthesised_speech.numpy() * 32767).astype(np.int16)
|
44 |
return 16000, synthesised_speech
|
@@ -46,9 +75,8 @@ def speech_to_speech_translation(audio):
|
|
46 |
|
47 |
title = "Cascaded STST"
|
48 |
description = """
|
49 |
-
Demo for cascaded speech-to-speech translation (STST), mapping from source speech in any language to target speech in
|
50 |
[SpeechT5 TTS](https://huggingface.co/microsoft/speecht5_tts) model for text-to-speech:
|
51 |
-
|
52 |
![Cascaded STST](https://huggingface.co/datasets/huggingface-course/audio-course-images/resolve/main/s2st_cascaded.png "Diagram of cascaded speech to speech translation")
|
53 |
"""
|
54 |
|
@@ -75,4 +103,4 @@ with demo:
|
|
75 |
gr.TabbedInterface([mic_translate, file_translate], ["Microphone", "Audio File"])
|
76 |
|
77 |
|
78 |
-
demo.launch()
|
|
|
2 |
import numpy as np
|
3 |
import torch
|
4 |
from datasets import load_dataset
|
5 |
+
import librosa
|
6 |
from transformers import SpeechT5ForTextToSpeech, SpeechT5HifiGan, SpeechT5Processor, pipeline
|
7 |
+
from transformers import WhisperProcessor, WhisperForConditionalGeneration
|
8 |
|
9 |
|
10 |
device = "cuda:0" if torch.cuda.is_available() else "cpu"
|
11 |
|
12 |
# load speech translation checkpoint
|
13 |
+
# asr_pipe = pipeline("automatic-speech-recognition", model="openai/whisper-base", device=device)
|
14 |
+
asr_processor = WhisperProcessor.from_pretrained("openai/whisper-base")
|
15 |
+
asr_model = WhisperForConditionalGeneration.from_pretrained("openai/whisper-base").to(device)
|
16 |
+
asr_forced_decoder_ids = asr_processor.get_decoder_prompt_ids(language="dutch", task="transcribe")
|
17 |
+
|
18 |
+
|
19 |
|
20 |
# load text-to-speech checkpoint and speaker embeddings
|
21 |
+
if 0:
|
22 |
+
processor = SpeechT5Processor.from_pretrained("sanchit-gandhi/speecht5_tts_vox_nl")
|
23 |
+
|
24 |
+
model = SpeechT5ForTextToSpeech.from_pretrained("sanchit-gandhi/speecht5_tts_vox_nl").to(device)
|
25 |
+
if 1:
|
26 |
+
from transformers import VitsModel, VitsTokenizer
|
27 |
+
model = VitsModel.from_pretrained("Matthijs/mms-tts-fra")
|
28 |
+
tokenizer = VitsTokenizer.from_pretrained("Matthijs/mms-tts-fra")
|
29 |
|
|
|
30 |
vocoder = SpeechT5HifiGan.from_pretrained("microsoft/speecht5_hifigan").to(device)
|
31 |
|
32 |
embeddings_dataset = load_dataset("Matthijs/cmu-arctic-xvectors", split="validation")
|
|
|
34 |
|
35 |
|
36 |
def translate(audio):
|
37 |
+
if 0:
|
38 |
+
outputs = asr_pipe(audio, max_new_tokens=256, generate_kwargs={"language":"dutch", "task":"transcribe"})
|
39 |
+
return outputs["text"]
|
40 |
+
else:
|
41 |
+
|
42 |
+
x, sr = librosa.load(audio)
|
43 |
+
input_features = asr_processor(x, sampling_rate=16000, return_tensors="pt").input_features
|
44 |
+
predicted_ids = asr_model.generate(input_features, forced_decoder_ids=asr_forced_decoder_ids)
|
45 |
+
# decode token ids to text
|
46 |
+
transcription = asr_processor.batch_decode(predicted_ids, skip_special_tokens=True)
|
47 |
+
return transcription
|
48 |
+
|
49 |
|
50 |
|
51 |
def synthesise(text):
|
52 |
+
if 0:
|
53 |
+
inputs = processor(text=text, return_tensors="pt")
|
54 |
+
speech = model.generate_speech(inputs["input_ids"].to(device), speaker_embeddings.to(device), vocoder=vocoder)
|
55 |
+
return speech.cpu()
|
56 |
+
if 1:
|
57 |
+
inputs = tokenizer(text, return_tensors="pt")
|
58 |
+
input_ids = inputs["input_ids"]
|
59 |
+
|
60 |
+
|
61 |
+
with torch.no_grad():
|
62 |
+
outputs = model(input_ids)
|
63 |
+
|
64 |
+
speech = outputs.audio[0]
|
65 |
+
return speech.cpu()
|
66 |
|
67 |
|
68 |
def speech_to_speech_translation(audio):
|
69 |
translated_text = translate(audio)
|
70 |
+
print(translated_text)
|
71 |
synthesised_speech = synthesise(translated_text)
|
72 |
synthesised_speech = (synthesised_speech.numpy() * 32767).astype(np.int16)
|
73 |
return 16000, synthesised_speech
|
|
|
75 |
|
76 |
title = "Cascaded STST"
|
77 |
description = """
|
78 |
+
Demo for cascaded speech-to-speech translation (STST), mapping from source speech in any language to target speech in Dutch. Demo uses OpenAI's [Whisper Base](https://huggingface.co/openai/whisper-base) model for speech translation, and Microsoft's
|
79 |
[SpeechT5 TTS](https://huggingface.co/microsoft/speecht5_tts) model for text-to-speech:
|
|
|
80 |
![Cascaded STST](https://huggingface.co/datasets/huggingface-course/audio-course-images/resolve/main/s2st_cascaded.png "Diagram of cascaded speech to speech translation")
|
81 |
"""
|
82 |
|
|
|
103 |
gr.TabbedInterface([mic_translate, file_translate], ["Microphone", "Audio File"])
|
104 |
|
105 |
|
106 |
+
demo.launch()
|