HarshBhanushali7705/F5-TTS-for-Gujarati-Language

🗣️ F5-TTS for Gujarati (24000 Iterations)

This is a fine-tuned F5-TTS model on the Gujarati language using the IIT Madras Indic TTS Dataset.

• —Base Model: SWivid/F5-TTS
• —Language: Gujarati (gu)
• —Training Steps: 24000
• —Sampling Rate: 22050 Hz
• —License: GPL-3.0

🧠 Model Details

This model is fine-tuned for Gujarati speech synthesis and is part of efforts to expand high-quality TTS to low-resource Indian languages. The model uses Tacotron-based architecture with attention and vocoder backend.

👥 Contributors

• —Harsh Bhanushali
• —Harsh Ahir (🤗 Hugging Face: Ahir4)

📦 Files

• —model_24000.pt – Fine-tuned model checkpoint
• —config.json – Model configuration (sampling rate, layers, etc.)
• —inference.py – Example inference script
• —assets/sample_output.wav – Optional audio sample

🔧 How to Use

Install dependencies

pip install torch torchaudio numpy