app.py

import gradio as gr
from transformers import AutoProcessor, BarkModel
import scipy.io.wavfile
import torch
import os

# Initialize model and processor
processor = AutoProcessor.from_pretrained("suno/bark")
model = BarkModel.from_pretrained("suno/bark")

def text_to_speech(text, voice_preset="v2/hi_speaker_2"):
    # Generate audio from text
    inputs = processor(text, voice_preset=voice_preset)
    
    # Generate audio
    audio_array = model.generate(**inputs)
    audio_array = audio_array.cpu().numpy().squeeze()
    
    # Get sample rate from model config
    sample_rate = model.generation_config.sample_rate
    
    # Create temporary file path
    output_path = "temp_audio.wav"
    
    # Save audio file
    scipy.io.wavfile.write(output_path, rate=sample_rate, data=audio_array)
    
    return output_path

# Define available voice presets
voice_presets = [
    "v2/hi_speaker_1",
    "v2/hi_speaker_2",
    "v2/hi_speaker_3",
    "v2/hi_speaker_4",
    "v2/hi_speaker_5"
]

# Create Gradio interface
demo = gr.Interface(
    fn=text_to_speech,
    inputs=[
        gr.Textbox(label="Enter text (Hindi or English)", placeholder="तुम बहुत अच्छे हो..."),
        gr.Dropdown(choices=voice_presets, value="v2/hi_speaker_2", label="Select Voice")
    ],
    outputs=gr.Audio(label="Generated Speech"),
    title="Bark Text-to-Speech",
    description="Convert text to speech using the Bark model. Supports Hindi and English text.",
    examples=[
        ["तुम बहुत अच्छे हो और मैं भी तुम्हारी तरह अच्छा हूँ", "v2/hi_speaker_2"],
        ["You are very nice and I am also nice like you", "v2/hi_speaker_1"]
    ]
)

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