Speech Models
Compare Voicetypr's speech recognition models. Size, speed, accuracy, and language support.
Compare Voicetypr's speech recognition models. Size, speed, accuracy, and language support.
Voicetypr runs speech recognition entirely on your device using open-source Whisper models.
| Model | Size | Speed | Accuracy | Best For |
|---|---|---|---|---|
| Whisper Base | ~140 MB | Fastest | Good | Clear speech, fast workflows |
| Whisper Large v3 | ~2.9 GB | Slower | Best | Noisy environments, accents |
| Whisper Large v3 Turbo | ~1.5 GB | Fast | Very good | Balance of speed and accuracy |
| Parakeet (Apple Silicon) | ~0.6 GB | Fast | Very good | Mac users, real-time use |
| Model | Minimum RAM | Recommended |
|---|---|---|
| Whisper Base | 4 GB | 8 GB |
| Whisper Large v3 Turbo | 8 GB | 16 GB |
| Whisper Large v3 | 16 GB | 32 GB |
| Parakeet | 4 GB | 8 GB |
All Whisper models support 99+ languages including:
English, Spanish, French, German, Italian, Portuguese, Dutch, Swedish, Polish, Japanese, Mandarin, Hindi, Arabic, Korean, Turkish, Russian, and 85 more.
Parakeet is optimized for English but supports other languages with varying accuracy.
Voicetypr can reformat your dictation with AI presets:
These run using your chosen AI provider (OpenAI, Anthropic, Google Gemini, or a custom OpenAI-compatible endpoint).