Products · Dictation
Kaho — Dictation that never leaves your side of the wire.
Kaho is a voice-dictation app for the desktop. Hold a hotkey, speak, release — cleaned, punctuated and, if you want, translated text appears wherever your cursor is. Speech-to-text runs on AI servers we own and operate, so the audio is never handed to a third-party inference API.
Kaho
Dictation · Shipping
What it does
Dictation everywhere you type: an editor, a browser field, a terminal, a chat window. Kaho does not ask you to work inside its own window — it puts finished text where your cursor already is. The transcript is cleaned before it lands: filler removed, punctuation and casing restored, and optionally translated into another language on the way.
Why it is built this way
Dictation is one of the most sensitive things software can touch — you speak client names, patient details, contract terms and passwords into it. Kaho therefore runs speech recognition on hardware we operate ourselves. There is no third-party inference API in the default data path, and no audio is kept after the text is returned. Cloud fallback exists, but it is a switch you turn on knowingly, not a default we hoped you would not audit.
The engineering behind it
Kaho is the reference implementation of our own AI & model-engineering practice: open speech models selected, quantized, benchmarked and served on our own GPU, behind a gateway that does the correction and translation passes. The same tuning work that gives Kaho its latency is the work we do for clients on their own hardware.
- macOS + Linux
- Runs on our own servers
- No audio retention
Where it lives