app.py

import gradio as gr
from transformers import SpeechT5ForTextToSpeech, SpeechT5Processor
import torch
import torchaudio
import tempfile

# Load model and processor
processor = SpeechT5Processor.from_pretrained("microsoft/speecht5_tts")
model = SpeechT5ForTextToSpeech.from_pretrained("microsoft/speecht5_tts")

# Load a voice embedding (necessary for the SpeechT5 model)
speaker_embedding, _ = torchaudio.load("https://huggingface.co/microsoft/speecht5_tts/blob/main/speaker_embeddings/english/vctk_speaker_0.pt")

def text_to_speech(text):
    inputs = processor(text, return_tensors="pt")
    speech = model.generate_speech(inputs["input_ids"], speaker_embedding)

    # Save the output to a temporary file
    with tempfile.NamedTemporaryFile(delete=False, suffix=".wav") as f:
        torchaudio.save(f.name, speech, 16000)
        return f.name

# Gradio interface
interface = gr.Interface(
    fn=text_to_speech,
    inputs="text",
    outputs="audio",
    title="Text to Speech",
    description="Convert text to speech using the microsoft/speecht5_tts model"
)

interface.launch()