Aizome sekka.computerSekka

built from d9f0217 ·

Dictation

Koe

Koe means voice.

Speak, and it's typed.

Generic name
Dictation
Command
aizome-koe
App ID
org.aizome.Koe
Koe's Dictation page: Dictation is off, with Download a Model and Turn On (Base, 148 MB); six speech models, each with Download; and the recognition options.
The installed Koe under a private headless Aizome compositor at 1280 × 800, captured with grim on 29 October 2026 for Sekka's application-startup check, with a fresh test profile and no models. The bar names its temporary desktop entry, Startup Window. 1280 × 800, not upscaled

Dictation

Press D in any app, speak, and press it again: what you said is typed where you were. Koe recognises speech on this computer with Whisper, on the graphics card through Vulkan where there is one.

Set up without a terminal

Download a Model and Turn On fetches Base (148 MB), checks it against a pinned SHA-256 and turns dictation on with its shortcut. Nothing is downloaded until you ask. Pick a model (Tiny, Base or Small, multilingual or English-only), the language, translation into English, the graphics card, and whether to finish after a pause.

Your words

Vocabulary and Teaching keeps packs of words, hints and approved corrections, shareable as .koe.json. Audio and transcripts stay in memory, and Koe checks the focus and the lock before it types.