app.py

import gradio as gr
import torch
from transformers import VitsModel, VitsTokenizer

# --- 1. Load Model and Tokenizer ---
# NOTE: Switched to a compatible model that has the correct file structure.
print("Loading facebook/mms-tts-eng model and tokenizer...")
model_id = "facebook/mms-tts-eng"
model = VitsModel.from_pretrained(model_id)
tokenizer = VitsTokenizer.from_pretrained(model_id)
print("Model and tokenizer loaded successfully.")

# --- 2. Define the Speech Synthesis Function ---
def synthesize_speech(text):
    """
    Converts text to speech using the selected TTS model.
    """
    # Tokenize the input text. The `return_tensors="pt"` part formats it for PyTorch.
    inputs = tokenizer(text, return_tensors="pt")

    # Generate the audio waveform.
    # We use torch.no_grad() to speed up inference as we aren't training the model.
    with torch.no_grad():
        waveform = model(**inputs).waveform

    # The output is a PyTorch tensor. Convert it to a NumPy array.
    # .squeeze() removes any extra single dimensions.
    waveform_numpy = waveform.cpu().numpy().squeeze()
    
    # Get the sampling rate from the model's configuration.
    sampling_rate = model.config.sampling_rate

    # Return the sampling rate and waveform as a tuple for the Gradio Audio component.
    return (sampling_rate, waveform_numpy)

# --- 3. Build the Gradio Interface ---
demo = gr.Interface(
    fn=synthesize_speech,
    inputs=gr.Textbox(
        label="Text to Synthesize",
        info="Enter the text you want to convert to speech.",
        value="Hello, this is a demonstration of the Facebook MMS text to speech model."
    ),
    outputs=gr.Audio(
        label="Synthesized Audio",
        type="numpy"  # The function returns a NumPy array
    ),
    title="🗣️ MMS Text-to-Speech (English)",
    description="A Gradio app to run the `facebook/mms-tts-eng` model for text-to-speech conversion.",
    examples=[
        ["The quick brown fox jumps over the lazy dog."],
        ["To be, or not to be, that is the question."],
        ["Artificial intelligence will shape our future in profound ways."]
    ],
    cache_examples=True # Cache results for faster demo
)

# --- 4. Launch the App ---
if __name__ == "__main__":
    demo.launch()