model.py

#!/usr/bin/env python3
import torch
from transformers import pipeline

device = "cuda:0" if torch.cuda.is_available() else "cpu"
pipe = pipeline(
    "automatic-speech-recognition", model="openai/whisper-base", device=device
)


from datasets import load_dataset

dataset = load_dataset("facebook/voxpopuli", "it", split="validation", streaming=True)
sample = next(iter(dataset))


from IPython.display import Audio

Audio(sample["audio"]["array"], rate=sample["audio"]["sampling_rate"])

def translate(audio):
    outputs = pipe(audio, max_new_tokens=256, generate_kwargs={"task": "translate"})
    return outputs["text"]

# print(translate(sample["audio"].copy()))
# print(sample["raw_text"])


# Text to speech

from transformers import SpeechT5Processor, SpeechT5ForTextToSpeech, SpeechT5HifiGan

processor = SpeechT5Processor.from_pretrained("microsoft/speecht5_tts")

model = SpeechT5ForTextToSpeech.from_pretrained("microsoft/speecht5_tts")
vocoder = SpeechT5HifiGan.from_pretrained("microsoft/speecht5_hifigan")

model.to(device)
vocoder.to(device)
embeddings_dataset = load_dataset("Matthijs/cmu-arctic-xvectors", split="validation")
speaker_embeddings = torch.tensor(embeddings_dataset[7306]["xvector"]).unsqueeze(0)

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()

speech = synthesise("Hey there! This is a test!")

Audio(speech, rate=16000)


# Concatenate the two models

import numpy as np

target_dtype = np.int16
max_range = np.iinfo(target_dtype).max


def speech_to_speech_translation(audio):
    translated_text = translate(audio)
    synthesised_speech = synthesise(translated_text)
    synthesised_speech = (synthesised_speech.numpy() * max_range).astype(np.int16)
    return 16000, synthesised_speech

sampling_rate, synthesised_speech = speech_to_speech_translation(sample["audio"])

Audio(synthesised_speech, rate=sampling_rate)