app.py

import gradio as gr
from transformers import SpeechT5Processor, SpeechT5ForTextToSpeech, SpeechT5HifiGan
from transformers import AutoProcessor, AutoModelForTextToSpectrogram
from datasets import load_dataset
import torch
import soundfile as sf
import os

# Load models and processors
processor = AutoProcessor.from_pretrained("ayush2607/speecht5_tts_technical_data")
model = AutoModelForTextToSpectrogram.from_pretrained("ayush2607/speecht5_tts_technical_data")
vocoder = SpeechT5HifiGan.from_pretrained("microsoft/speecht5_hifigan")

# Load xvector containing speaker's voice characteristics from a dataset
embeddings_dataset = load_dataset("Matthijs/cmu-arctic-xvectors", split="validation")
speaker_embeddings = torch.tensor(embeddings_dataset[7306]["xvector"]).unsqueeze(0)

def text_to_speech(text):
    inputs = processor(text=text, return_tensors="pt")
    speech = model.generate_speech(inputs["input_ids"], speaker_embeddings, vocoder=vocoder)
    
    output_path = "output.wav"
    sf.write(output_path, speech.numpy(), samplerate=16000)
    
    return output_path

# Create Gradio interface
iface = gr.Interface(
    fn=text_to_speech,
    inputs=gr.Textbox(label="Enter text to convert to speech"),
    outputs=gr.Audio(label="Generated Speech"),
    title="Text-to-Speech Converter",
    description="Convert text to speech using the SpeechT5 model."
)

# Launch the app
iface.launch()