Troubleshooting › Fix: Whisper Accent Transcription Issues
Symptom: Whisper consistently mis-transcribes your speech. Common words are recognized incorrectly. Technical terms produce gibberish.
⚠ Base model too small
→ The base model (142MB) has limited accuracy. Upgrade to small (466MB) or medium (1.5GB). Run MODEL=small bash scripts/install-whisper.sh.
⚠ Background noise
→ Use a headset mic or move to a quieter environment. Background noise significantly reduces whisper accuracy.
⚠ Speaking too fast
→ Speak at a moderate pace. Whisper works best with clear, natural speech. Pause briefly between sentences.
⚠ Technical vocabulary
→ Set OPENAI_API_KEY. OpenAI's Whisper API handles code terms, APIs, and technical jargon much better than local models.
⚠ Wrong model for your language
→ Whisper supports 100+ languages. If English is not your first language, speaking English with an accent works better with larger models.
Set OPENAI_API_KEY as your primary transcription method. The plugin tries OpenAI API first when the key is available. OpenAI handles accents significantly better than any local whisper model.
Still having trouble?
Contact support →