Setting Up AI Enhancement with Your Own API Key
A quick guide to connecting OpenAI, Anthropic, or any compatible provider for transcript cleanup.
Whiskers transcribes your speech on-device, but the raw transcript can benefit from a cleanup pass—fixing grammar, adding punctuation, and polishing phrasing. That's where AI Enhancement comes in.
What AI Enhancement does
After Whiskers generates a raw transcript, AI Enhancement sends that text (not your audio) to an LLM of your choice for cleanup. The result is a polished version that reads like something you'd actually write.
Connecting your API key
- Open Whiskers Settings and navigate to the AI Enhancement tab.
- Select your provider: OpenAI, Anthropic, or any OpenAI-compatible endpoint.
- Paste your API key into the field provided.
- Choose a model. We recommend GPT-4o or Claude 3.5 Sonnet for the best balance of quality and speed.
- Click Test Connection to verify everything works.
That's it. From now on, every transcript will be automatically enhanced before it's pasted.
Privacy considerations
AI Enhancement sends your transcript text to a third-party API. If this conflicts with your privacy requirements, you can leave it disabled—Whiskers works perfectly well with just on-device transcription. The enhanced text is never stored by Whiskers; it's fetched, applied, and discarded.
Tips for best results
- Use a capable model. Smaller models tend to over-correct or miss context.
- Keep transcripts reasonably sized. Very long dictation sessions may hit token limits. Consider breaking them into segments.
- Review the output. AI Enhancement is good, but it's not perfect. A quick scan catches the occasional hallucination.