app.py

# app.py
import gradio as gr
from transformers import pipeline

# -------------------------------
# Define your TTS models here
# -------------------------------
# Add model names from Hugging Face Hub
tts_models = {
    "Indic Parler TTS (Indian English)": "ai4bharat/indic-parler-tts",
    # "Hindi TTS (FuturixAI)": "Futurix-AI/Hindi-TTS",
    # "Hinglish TTS (sachin2000keshav)": "sachin2000keshav/hinglish_tts_v1",
    # "MMS-TTS Hindi": "facebook/mms-tts-hin"
}

# Load pipelines
pipelines = {}
for name, model_name in tts_models.items():
    print(f"Loading model: {name}")
    pipelines[name] = pipeline("text-to-speech", model=model_name)

# -------------------------------
# Function to run TTS
# -------------------------------
def generate_speech(text, model_choice):
    if model_choice not in pipelines:
        return None
    tts_pipeline = pipelines[model_choice]
    # Generate speech
    audio = tts_pipeline(text)
    # The output format may vary; handle bytes or dict
    if isinstance(audio, dict) and "audio" in audio:
        return audio["audio"]
    return audio

# -------------------------------
# Gradio UI
# -------------------------------
with gr.Blocks() as demo:
    gr.Markdown("## Test Multiple Indian TTS Models")
    text_input = gr.Textbox(label="Enter text", placeholder="Type text in Hindi, Hinglish, or English", lines=3)
    model_dropdown = gr.Dropdown(label="Select TTS Model", choices=list(tts_models.keys()))
    output_audio = gr.Audio(label="Generated Speech")
    
    text_input.submit(generate_speech, inputs=[text_input, model_dropdown], outputs=output_audio)
    model_dropdown.change(generate_speech, inputs=[text_input, model_dropdown], outputs=output_audio)

demo.launch()