<p align="center">
  <img src="./assets/logo.png" alt="pi-vosk-voice logo" width="220" />
</p>

# pi-vosk-voice

Push-to-talk voice input for Pi using Vosk offline speech recognition.

## Features

- Hold-to-talk recording
- Live transcription while speaking
- Partial stabilization on pauses
- Writes into the Pi editor in `paste` mode
- Optional `send` mode
- Start/stop audio cues via SoX
- Global or project-local Pi configuration

## Requirements

- `sox` (`rec` and `play`)
- Python 3
- `vosk` Python package
- A downloaded Vosk model, e.g. `vosk-model-it-0.22`

## Configuration

Put this in `~/.pi/vosk-voice.json` or `.pi/vosk-voice.json`:

```json
{
  "shortcut": "f12",
  "enabled": true,
  "mode": "paste",
  "modelPath": "/home/you/.cache/vosk/vosk-model-it-0.22",
  "soundsEnabled": true
}
```

## Install in Pi

```bash
pi install /path/to/pi-vosk-voice
```

Then reload Pi:

```text
/reload
```

## Commands

- `/voice`
- `/voice config`
- `/voice stop`
- `/voice download-model`
- `/voice set shortcut ctrl+t`
- `/voice set mode paste`
- `/voice set sounds false`

## Notes

The extension uses Vosk streaming partial/final results and updates the editor progressively.
## Logo Variants

### Smooth / vector-like

![Smooth logo](./assets/variants/logo_smooth.png)

### Terminal / hacker style

![Hacker logo](./assets/variants/logo_hacker.png)