Spaces:
Runtime error
Runtime error
File size: 2,822 Bytes
e59aa55 359a43b e59aa55 359a43b e59aa55 359a43b e59aa55 359a43b e59aa55 359a43b e59aa55 eff8c09 e59aa55 |
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 |
import gradio as gr
import numpy as np
import torch
from datasets import load_dataset
from transformers import SpeechT5ForTextToSpeech, SpeechT5HifiGan, SpeechT5Processor, pipeline
device = "cuda:0" if torch.cuda.is_available() else "cpu"
# load speech translation checkpoint
asr_pipe = pipeline("automatic-speech-recognition", model="WasuratS/whisper-base-danish-finetune-common-voice-11", device=device)
# load text-to-speech checkpoint and speaker embeddings
processor = SpeechT5Processor.from_pretrained("WasuratS/speecht5_finetuned_voxpopuli_nl")
model = SpeechT5ForTextToSpeech.from_pretrained("WasuratS/speecht5_finetuned_voxpopuli_nl").to(device)
vocoder = SpeechT5HifiGan.from_pretrained("microsoft/speecht5_hifigan").to(device)
embeddings_dataset = load_dataset("Matthijs/cmu-arctic-xvectors", split="validation")
speaker_embeddings = torch.tensor(embeddings_dataset[7306]["xvector"]).unsqueeze(0)
def translate(audio):
outputs = asr_pipe(audio, max_new_tokens=256, generate_kwargs={"task": "translate", "language": "nl"})
return outputs["text"]
def synthesise(text):
inputs = processor(text=text, return_tensors="pt")
speech = model.generate_speech(inputs["input_ids"].to(device), speaker_embeddings.to(device), vocoder=vocoder)
return speech.cpu()
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
title = "Cascaded STST - Danish to Dutch"
description = """
Demo for cascaded speech-to-speech translation (STST), mapping from source speech in Danish language to target speech in Dutch ! <br/> Demo uses my fine tuned OpenAI's [Whisper Base](WasuratS/whisper-base-danish-finetune-common-voice-11) model for speech translation, and my fine tuned Microsoft's
[SpeechT5 TTS](https://huggingface.co/WasuratS/speecht5_finetuned_voxpopuli_nl) model 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()
|