Models
Choosing a Model
Compare all available transcription models in Vorn Voice to find the right one for you.
Vorn Voice includes six local models and two cloud options. Here is how they compare.
Local models
| Model | Based on | Size | Speed | Accuracy | Best for |
|---|---|---|---|---|---|
| Vorn Lite | Moonshine Tiny | ~118 MB | Fastest | Good | Quick messages, low-spec hardware |
| Vorn Standard | Moonshine Base | ~274 MB | Fast | Better | Most users (default) |
| Vorn Compact | Whisper Tiny | ~103 MB | Fast | Good | Small download, basic dictation |
| Vorn Balanced | Whisper Base | ~160 MB | Medium | Better | Balance of speed and quality |
| Vorn Pro | Whisper Small | ~374 MB | Slower | High | Longer dictations, fewer errors |
| Vorn Max | Whisper Medium | ~945 MB | Slowest | Highest | Maximum local accuracy |
Recommendations
- Start with Vorn Standard. It is the default for a reason — fast enough for push-to-talk and accurate enough for everyday dictation.
- Try Vorn Lite if you want the fastest possible response time and can tolerate occasional small errors.
- Try Vorn Pro or Vorn Max if you regularly dictate longer passages and want fewer corrections.
Cloud models
| Model | Provider | Speed | Accuracy | Languages |
|---|---|---|---|---|
| Vorn Cloud | Groq (proxy) | Fast (network dependent) | High | Many |
| BYOK Groq | Groq (direct) | Fast (network dependent) | High | Many |
Cloud transcription uses Groq Whisper Large V3 Turbo, which is more accurate than any local model and supports dozens of languages. The tradeoff is that it requires internet and sends audio to the server.
Switching models
Open settings and go to the model section. Select a model from the list. If it is not yet downloaded, the app will download it automatically. Your previous local model is remembered so you can switch back quickly.