Settings

The Whisper model used for transcription. Accuracy improves from tiny → base → small → medium, but processing time increases. Uninstalled models can be downloaded here.

The language used for speech recognition. Choose from auto (detect) or one of 13 supported languages: Japanese, English, Chinese, Korean, French, German, Spanish, Portuguese, Russian, Arabic, Hindi, and Italian.

The keyboard shortcut to start/stop voice input.

After recognition, a local LLM removes filler words and applies formatting. Custom and app-specific instructions can be configured in the LLM tab.

When enabled, the transcription result is also copied to the clipboard after each recognition. If auto-insertion fails, you can still paste manually with ⌘V.