> ## Documentation Index
> Fetch the complete documentation index at: https://docs.openwhispr.com/llms.txt
> Use this file to discover all available pages before exploring further.

# Local models

> Set up Whisper and NVIDIA Parakeet for private, offline transcription.

Local processing keeps your audio on your device. OpenWhispr supports two speech recognition engines.

## OpenAI Whisper (via whisper.cpp)

The default local engine. Choose a model based on your needs:

| Model  | Size     | Speed    | Quality            |
| ------ | -------- | -------- | ------------------ |
| tiny   | \~75 MB  | Fastest  | Basic              |
| base   | \~142 MB | Fast     | Good (recommended) |
| small  | \~466 MB | Moderate | Better             |
| medium | \~1.5 GB | Slow     | High               |
| large  | \~3 GB   | Slowest  | Best               |
| turbo  | \~1.6 GB | Fast     | Good               |

### Setup

1. Go to **Settings > Speech Recognition > Local**
2. Select a Whisper model
3. Click **Download** — models are stored in `~/.cache/openwhispr/whisper-models/`

The whisper.cpp binary is bundled with OpenWhispr. No Python or additional runtime needed.

## NVIDIA Parakeet (via sherpa-onnx)

A faster alternative to Whisper, especially on lower-end hardware.

| Model                    | Size     | Languages                           |
| ------------------------ | -------- | ----------------------------------- |
| parakeet-tdt-0.6b-v3     | \~680 MB | 25 languages                        |
| parakeet-unified-en-0.6b | \~631 MB | English (state-of-the-art accuracy) |

### Setup

1. Go to **Settings > Speech Recognition > Local**
2. Switch the provider to **NVIDIA Parakeet**
3. Download the model

Parakeet uses INT8 quantized ONNX models for efficient CPU inference.

## Which to choose

* **Parakeet** — best for speed, lower-end hardware, or when you need multilingual support
* **Whisper** — best for quality, especially with the `medium` or `large` models

## Disk management

Remove downloaded models anytime from **Settings > General > Local Model Storage > Remove Downloaded Models**. Models can be re-downloaded when needed.