app.py

import gradio as gr
from kittentts import KittenTTS
import soundfile as sf
import os

# 1. Initialize the model
# Using "." tells the library to look for the .onnx, .npz, and .json files 
# you just uploaded to the local directory.
# (If it throws an error, you can safely change "." to "KittenML/kitten-tts-mini-0.8" 
# and it will still work perfectly inside your Space).
try:
    model = KittenTTS(".")
except Exception:
    model = KittenTTS("KittenML/kitten-tts-mini-0.8")

# Available voices in version 0.8
AVAILABLE_VOICES =['Bella', 'Jasper', 'Luna', 'Bruno', 'Rosie', 'Hugo', 'Kiki', 'Leo']

# 2. Define the synthesis function
def synthesize_audio(text, voice):
    # Generate the raw audio array
    audio = model.generate(text, voice=voice)
    
    # Save it to a .wav file so Gradio can play it
    output_path = "output.wav"
    sf.write(output_path, audio, 24000)
    
    return output_path

# 3. Build the Gradio UI
with gr.Blocks(title="Kitten TTS Mini") as demo:
    gr.Markdown("# 🐱 Kitten TTS Mini 0.8")
    gr.Markdown("Type some text below to generate speech using your uploaded ONNX model!")
    
    with gr.Row():
        with gr.Column():
            text_input = gr.Textbox(label="Text to Speak", placeholder="Hello, I am a lightweight AI...", lines=4)
            voice_dropdown = gr.Dropdown(choices=AVAILABLE_VOICES, value="Jasper", label="Select Voice")
            generate_btn = gr.Button("Generate Speech", variant="primary")
            
        with gr.Column():
            audio_output = gr.Audio(label="Generated Audio", type="filepath")
            
    # Connect the button to the function
    generate_btn.click(
        fn=synthesize_audio, 
        inputs=[text_input, voice_dropdown], 
        outputs=audio_output
    )

# Launch the app
if __name__ == "__main__":
    demo.launch()