Spaces:
Paused
Paused
| import torch | |
| import numpy as np | |
| from transformers import pipeline | |
| from transformers import BarkModel | |
| from transformers import AutoProcessor | |
| device="cpu" | |
| pipe = pipeline( | |
| "automatic-speech-recognition", model="openai/whisper-large-v2", device=device | |
| ) | |
| processor = AutoProcessor.from_pretrained("suno/bark") | |
| model = BarkModel.from_pretrained("suno/bark") | |
| model = model.to(device) | |
| synthesised_rate = model.generation_config.sample_rate | |
| def translate(audio): | |
| outputs = pipe(audio, max_new_tokens=256, generate_kwargs={"task": "transcribe","language":"chinese"}) | |
| return outputs["text"] | |
| def synthesise(text_prompt,voice_preset="v2/zh_speaker_1"): | |
| inputs = processor(text_prompt, voice_preset=voice_preset) | |
| speech_output = model.generate(**inputs.to(device),pad_token_id=10000) | |
| return speech_output | |
| def speech_to_speech_translation(audio): | |
| translated_text = translate(audio) | |
| synthesised_speech = synthesise(translated_text) | |
| synthesised_speech = (synthesised_speech.numpy() * 32767).astype(np.int16) | |
| return 16000, synthesised_speech | |
| def speech_to_speech_translation(audio,voice_preset="v2/zh_speaker_1"): | |
| translated_text = translate(audio) | |
| synthesised_speech = synthesise(translated_text,voice_preset) | |
| synthesised_speech = (synthesised_speech.numpy() * 32767).astype(np.int16) | |
| return synthesised_rate , synthesised_speech | |
| def speech_to_speech_translation_fix(audio,voice_preset="v2/zh_speaker_1"): | |
| synthesised_rate,synthesised_speech = speech_to_speech_translation(audio,voice_preset) | |
| return synthesised_rate,synthesised_speech.T | |
| title = "Multilanguage to Chinese(mandarin) Cascaded STST" | |
| description = """ | |
| Demo for cascaded speech-to-speech translation (STST), mapping from source speech in Multilanguage to target speech in Chinese(mandarin). Demo uses OpenAI's [Whisper arge-v2](https://huggingface.co/openai/whisper-large-v2) model for speech translation, and a suno/bark[bark-small](https://huggingface.co/suno/bark) model for text-to-speech: | |
|  | |
| """ | |
| examples = [ | |
| ["./mama (1).mp3", None], | |
| ["./mama (2).mp3", None], | |
| ["./mama (3).mp3", None], | |
| ["./mama (4).mp3", None], | |
| ["./mama (5).mp3", None], | |
| ["./mama (6).mp3", None], | |
| ["./mama (7).mp3", None], | |
| ["./mama (8).mp3", None], | |
| ] | |
| import gradio as gr | |
| demo = gr.Blocks() | |
| # Muti_translate=gr.Interface( | |
| # fn=speech_to_speech_translation_fix, | |
| # inputs=[ | |
| # gr.Audio(label="Upload Speech", source="upload", type="filepath"), | |
| # gr.Audio(label="Record Speech", source="microphone", type="filepath"), | |
| # ], | |
| # outputs=[ | |
| # gr.Audio(label="Generated Speech", type="numpy"), | |
| # gr.Text(label="Transcription"), | |
| # ], | |
| # title=title, | |
| # description=description, | |
| # examples=examples, | |
| # ) | |
| mic_translate = gr.Interface( | |
| fn=speech_to_speech_translation_fix, | |
| inputs=gr.Audio(source="microphone", type="filepath"), | |
| outputs=gr.Audio(label="Generated Speech", type="numpy"), | |
| title=title, | |
| description=description, | |
| ) | |
| file_translate = gr.Interface( | |
| fn=speech_to_speech_translation_fix, | |
| inputs=gr.Audio(source="upload", type="filepath"), | |
| outputs=gr.Audio(label="Generated Speech", type="numpy"), | |
| examples=examples, | |
| title=title, | |
| description=description, | |
| ) | |
| with demo: | |
| gr.TabbedInterface([mic_translate, file_translate], ["Microphone", "Audio File"]) | |
| #gr.TabbedInterface([Muti_translate], ["Record or upload your speech"]) | |
| demo.launch(share=True) |