---
title: LinaCodec Audio Tool
emoji: 🎵
colorFrom: blue
colorTo: purple
sdk: gradio
sdk_version: 4.44.0
app_file: app.py
pinned: false
license: mit
---

# 🎵 LinaCodec Audio Tool

A Gradio web interface for **LinaCodec**, a neural audio codec for high-quality speech compression and voice conversion.

## Features

### 🔄 Encode & Decode
- Upload audio files and compress them into speech tokens
- Decode tokens back to 48kHz audio
- Demonstrates codec's compression and reconstruction quality

### 🎭 Voice Conversion
- Transfer timbre/style from one speaker to another
- Keep content from source audio
- Apply voice characteristics from reference audio

## How to Use

1. **Encode & Decode Tab**:
   - Upload an audio file or record using your microphone
   - Click "Encode & Decode"
   - Listen to the reconstructed audio

2. **Voice Conversion Tab**:
   - Upload source audio (the content you want)
   - Upload reference audio (the voice style you want)
   - Click "Convert Voice"
   - Listen to the converted result

## Technical Details

- **Model**: [YatharthS/LinaCodec](https://huggingface.co/YatharthS/LinaCodec)
- **Output Sample Rate**: 48 kHz
- **Input Formats**: WAV, MP3, and other common audio formats
- **Framework**: PyTorch, Gradio

## Model Information

LinaCodec is a neural audio codec designed for high-quality speech compression and voice conversion. It encodes audio into discrete speech tokens and a global embedding, enabling efficient storage and manipulation of speech signals.

## Local Development

```bash
# Clone the repository
git clone  https://huggingface.co/spaces/humair025/LinaCodec/

# Install dependencies
pip install -r requirements.txt

# Run the app
python app.py
```

## Citation

If you use LinaCodec in your research, please cite the original work.

## License

MIT License