Spaces:
Sleeping
Sleeping
File size: 2,516 Bytes
1c2b467 7beb980 1c2b467 7beb980 1c2b467 7beb980 1c2b467 602984b 1c2b467 7beb980 1c2b467 7beb980 1c2b467 7beb980 1c2b467 7beb980 1c2b467 7beb980 602984b 1c2b467 7beb980 1c2b467 7beb980 1c2b467 7beb980 |
1 2 3 4 5 6 7 8 9 10 11 12 13 14 15 16 17 18 19 20 21 22 23 24 25 26 27 28 29 30 31 32 33 34 35 36 37 38 39 40 41 42 43 44 45 46 47 48 49 50 51 52 53 54 55 56 57 58 59 60 61 62 63 64 65 66 67 68 69 70 71 72 73 74 75 76 |
import torch
from transformers import pipeline, VitsModel, VitsTokenizer
import numpy as np
import gradio as gr
device = "cuda:0" if torch.cuda.is_available() else "cpu"
# Load Whisper-base
pipe = pipeline("automatic-speech-recognition",
model="openai/whisper-base",
device=device
)
# Load the model checkpoint and tokenizer
model = VitsModel.from_pretrained("Matthijs/mms-tts-fra")
tokenizer = VitsTokenizer.from_pretrained("Matthijs/mms-tts-fra")
# Define a function to translate an audio, in French here
def translate(audio):
outputs = pipe(audio, max_new_tokens=256,
generate_kwargs={"task": "transcribe", "language": "fr"})
return outputs["text"]
# Define function to generate the waveform output
def synthesise(text):
inputs = tokenizer(text, return_tensors="pt")
input_ids = inputs["input_ids"]
with torch.no_grad():
outputs = model(input_ids)
return outputs.audio[0]
# Define the pipeline
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
# Define the title etc
title = "Cascaded STST"
description = """
Demo for cascaded speech-to-speech translation (STST), mapping from source speech in any language to target speech in French. Demo uses OpenAI's [Whisper Base](https://huggingface.co/openai/whisper-base) model for speech translation, and Facebook's
[MMS TTS](https://huggingface.co/facebook/mms-tts) model, finetuned by [Matthijs](https://huggingface.co/Matthijs), for text-to-speech:
"""
demo = gr.Blocks()
mic_translate = gr.Interface(
fn=speech_to_speech_translation,
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,
inputs=gr.Audio(source="upload", type="filepath"),
outputs=gr.Audio(label="Generated Speech", type="numpy"),
examples=[["./example.wav"]],
title=title,
description=description,
)
with demo:
gr.TabbedInterface([mic_translate, file_translate], ["Microphone", "Audio File"])
demo.launch() |