Languages

Multi-language transcription support in Vowen.

Supported Languages

Vowen supports 99 languages via Whisper models and 25 European languages via Parakeet.

Setting Your Language

Go to Settings > Language and select your language from the dropdown.

Auto-Detect vs Fixed Language

Mode	When to Use
Auto-detect	You speak multiple languages and want Vowen to figure it out
Fixed language	You primarily speak one language (more accurate)

For best accuracy, set a fixed language instead of relying on auto-detect. Auto-detect can occasionally switch to the wrong language mid-sentence.

Pro tip: Switching languages mid-sentence? Set your language to Auto-detect and pair it with Soniox, which handles real-time multilingual transcription (English to Hindi, Spanish to English, and so on) better than anything else we’ve tested. Soniox is a paid streaming model, but on our two-person team, $5 of credit has lasted more than six months of daily use.

Translate to English

Enable Settings > Language > Translate to English to automatically translate non-English speech into English text. This uses Whisper’s built-in translation capability.

Language-Specific Recommendations

Language	Recommended Model	Notes
English	Base.en or Parakeet	.en models optimized for English
Spanish	Multilingual Small or Groq	Works well with local models
French	Large v3 Turbo or Groq	Good accuracy with larger models
German	Parakeet or Large v3	Parakeet supports German natively
Swedish	Large v3 Turbo or Groq Large v3	Smaller models struggle with Swedish
Romanian	Groq Whisper Large v3	Cloud recommended for best results
Indian languages	Sarvam AI Saaras v3	Specialized for Hindi, Tamil, Telugu, etc.
Japanese/Chinese	Large v3 or Groq	Needs larger models for accuracy
Arabic/Hebrew	Large v3 or cloud models	RTL languages supported

Parakeet Language Support

Parakeet TDT 0.6B v3 supports these 25 languages with auto-detection:

Bulgarian, Croatian, Czech, Danish, Dutch, English, Estonian, Finnish, French, German, Greek, Hungarian, Italian, Latvian, Lithuanian, Maltese, Polish, Portuguese, Romanian, Russian, Slovak, Slovenian, Spanish, Swedish, Ukrainian.

Tips for Non-English Transcription

1. Use a multilingual model — avoid .en models which are English-only
2. Set language explicitly — don’t rely on auto-detect for non-English
3. Try cloud models — Groq Whisper Large v3 often gives better non-English results than local models
4. Add vocabulary — Add names and technical terms specific to your language in Settings > Dictionary