Voice dictation
Dictate into the apps where you already type. Set your shortcut, choose how much Kimiko tidies your words, and add names to the dictionary.
Set up your shortcut
Section titled “Set up your shortcut”- Open Settings → Dictation and enable Dictate anywhere (Flow).
- On Mac, grant Accessibility access when prompted.
- In the Shortcuts card, check or record your Dictation key.
- Click into a text field in another app.
- Use your configured shortcut to speak, then stop to insert the text.
The current default is Ctrl on Mac and Ctrl + Win on Windows. Your saved shortcut may differ; the one shown in Settings is authoritative.
A held key records while you hold it and inserts when you release it. A key combination can use Hold to talk or Toggle behavior. With a held-key binding, optional Hands-free mode lets you double-tap to keep dictating and press again to stop.
Choose how much to tidy
Section titled “Choose how much to tidy”Under Behavior → Formatting, choose the result you want:
| Option | Result |
|---|---|
| Off | The speech model’s transcript, without extra cleanup |
| Clean | Lightweight cleanup without an additional AI pass |
| Polish | AI-assisted grammar and punctuation |
| Context-aware | AI formatting that adapts to the destination app |
For the quickest starting point, try Clean. For polished prose, try Polish and review the result. The formatting model is a separate choice from the speech model.
Models and language
Section titled “Models and language”Dictation has its own Speech model and Language settings. Changing your meeting transcription model does not automatically change dictation.
The Fastest option uses Parakeet v3. Check language support in the picker before choosing another model. Optional cloud choices send dictation audio to the selected provider.
Names and specialized words
Section titled “Names and specialized words”Add names, product terms, and jargon under Dictionary. If formatting changes your wording, Last dictation lets you copy the raw transcript of your most recent dictation.
Memory between dictations
Section titled “Memory between dictations”Since Kimiko 0.6.7, Settings → Dictation → Keep models loaded controls how long dictation’s speech and built-in formatting models stay ready after use. The default is 5 minutes. Options run from Unload right after each dictation to Never unload.
A shorter window frees memory sooner; the next dictation may take a little longer to start. A longer window keeps models ready but uses memory between dictations. Kimiko starts preparing the built-in formatting model when dictation begins, when that model is enabled and selected. A model still needed for another job, such as chat, can stay loaded longer.
See system requirements for hardware recommendations and memory use on laptops for shared-memory guidance.
If text does not appear
Section titled “If text does not appear”Keep the destination field focused. Password fields and some protected or elevated windows may prevent insertion. Finish an active meeting recording before testing dictation, since both need the microphone.
Check permissions and the dictation troubleshooting steps.