README.md

metadata

title: Vakya 2.0 - Text-to-Speech
emoji: 🎙️
colorFrom: blue
colorTo: purple
sdk: gradio
sdk_version: 6.2.0
app_file: app.py
pinned: false
license: mit

🎙️ Vakya 2.0 - Text-to-Speech Playground

Vakya is a high-quality Text-to-Speech model based on the IndicF5 architecture, supporting 11 Indian languages.

🌟 Features

• Multi-language Support: Assamese, Bengali, Gujarati, Hindi, Kannada, Malayalam, Marathi, Odia, Punjabi, Tamil, Telugu
• Voice Cloning: Uses reference audio to clone voice characteristics
• High Quality: 24kHz sample rate, 0.4B parameter model
• Easy to Use: Simple interface for testing and experimentation

🚀 How to Use

1. Load Model: Click the "Load Model" button (first time may take a few minutes to download)
2. Upload Reference Audio: Upload a short audio clip (<15 seconds recommended) that represents the voice you want to clone
3. Enter Reference Text (Optional): Type what is spoken in the reference audio. If left blank, the model will auto-transcribe it
4. Enter Text to Generate: Type the text you want to synthesize in any supported language
5. Adjust Settings (Optional):
   • Speed: Control the speech rate (0.5x to 2.0x)
   • Remove Silences: Experimental feature to remove pauses
6. Generate: Click "Generate Speech" and wait for the audio output

📋 Model Information

• Model: Vakya 2.0
• Repository: ashishkblink/vakya2.0
• Based on: IndicF5 by AI4Bharat (IIT Madras)
• Model Size: 0.4B parameters
• Sample Rate: 24000 Hz
• Training Data: 1417 hours of high-quality speech
• License: MIT License

💡 Tips for Best Results

• Keep reference audio clips short (<15 seconds) for best results
• Use clear, high-quality reference audio
• Provide reference text when possible for better voice matching
• The model works best with native speakers of the target language

⚠️ Terms of Use

• You must have explicit permission to clone voices
• Unauthorized voice cloning is strictly prohibited
• Any misuse of this model is the responsibility of the user
• This model is for research and educational purposes

🔗 Links

• Model Repository: ashishkblink/vakya2.0
• GitHub: ashishkblink/vakya
• IndicF5: AI4Bharat/IndicF5

🙏 Acknowledgments

This model is based on IndicF5 developed by AI4Bharat (IIT Madras).

---

Vakya - Bringing voices to Indian languages 🎙️