Rinari docsWorking with Rinari
Dictation
Talk to Rinari instead of typing. Dictation runs on your computer with whisper.cpp — no audio is sent to any service.
You are here
You can speak a request instead of typing it. Dictation is transcribed on your computer with whisper.cpp: the audio never leaves your machine, and it’s deleted once it’s transcribed.
Dictate
- Press the microphone in the composer to start, and press it again to stop.
- Or, in the message box, hold Ctrl+Space while you talk and release it to transcribe.
- Esc cancels a recording.
The transcribed text goes into the message box so you can review it and press Enter. One dictation takes up to 5 minutes.
The first time
Dictation needs a speech model, downloaded once and kept on your computer. The first time you dictate, Rinari tells you the model is missing and how big it is; after that download, dictation works offline.
Settings → Dictation
| Setting | What it does |
|---|---|
| Model | Three sizes: a light one for modest computers, the recommended one, and the most accurate, best with a GPU. Larger is more accurate and slower without a GPU. Download or remove each one here. |
| Language | The app’s language, or Detect automatically (slower). |
| Vocabulary | Names and terms to spell right, separated by commas. «Rinari» is already included. |
| Send when dictation ends | Off by default: the text stays in the message for you to review. On: it’s sent as soon as dictation ends. |
If it doesn’t work
- Rinari is not allowed to use the microphone — allow it in Windows’ microphone privacy settings.
- Nothing was understood — speak closer to the microphone.
- Transcription took too long — try a lighter model.