from transformers import pipeline
import torch
from transformers import SpeechT5Processor, SpeechT5ForTextToSpeech, SpeechT5HifiGan
from datasets import load_dataset
from IPython.display import Audio
device = "cuda" if torch.cuda.is_available() else "cpu"
from transformers import pipeline

pipe = pipeline(
    "automatic-speech-recognition", model="openai/whisper-base", device=device
)

dataset = load_dataset("facebook/voxpopuli", "es", split="validation", streaming=True,trust_remote_code=True)
sample = next(iter(dataset))
def translate(audio):
    outputs = pipe(audio, generate_kwargs={"task": "translate",'max_new_tokens':255})
    return outputs["text"]
# Load SpeechT5 Processor and Model
processor = SpeechT5Processor.from_pretrained("microsoft/speecht5_tts")
model = SpeechT5ForTextToSpeech.from_pretrained("microsoft/speecht5_tts").to(device)
vocoder = SpeechT5HifiGan.from_pretrained("microsoft/speecht5_hifigan").to(device)

# Load Speaker Embeddings
embeddings_dataset = load_dataset("Matthijs/cmu-arctic-xvectors", split="validation")
speaker_embeddings = torch.tensor(embeddings_dataset[0]["xvector"]).unsqueeze(0).to(device)
# Define Speech Synthesis Function
def synthesise(text):
    inputs = processor(text=text, return_tensors="pt")
    speech = model.generate_speech(
        inputs["input_ids"].to(device),
        speaker_embeddings,
        vocoder=vocoder
    )
    return speech.cpu()
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"])