Hold a key, talk the way you actually talk — fillers, false starts and all — then let go. sayso’s on-device AI cleans it up and drops finished text right at your cursor, in any app. No cloud. No account. No subscription.
Four things a plain dictation tool won’t do for you — all running on your own silicon.
Say a trigger word, paste a whole block. Addresses, standups, signatures, boilerplate replies.
Formal in Outlook, loose in Discord, terse in your terminal. Same voice, right register — sayso knows which app has focus.
Think in your language, write in theirs. Speech goes in however it comes out of you; text lands in the language you picked.
Whisper transcribes, Qwen3 cleans up — both running locally through whisper.cpp and llama.cpp. GPU when you have one, CPU when you don’t.
Streaks, words dictated, time saved, per-app usage — every number is computed on this computer and never leaves it. Switch the range below to replay your last weeks of activity.
This is the real app, running right here.
Click through every screen. Open Appearance to switch between seven themes and dark, light or auto, drop in a living background that follows your cursor, or pick a sound-wave style. Every change repaints the whole window instantly — no reload.
The same words, retuned to the register you need — never the meaning. Pick a default tone, or set a different one per app.
The same “edits, not just transcribes” magic you’d expect from a cloud app — without the cloud, the account, or the bill.
MIT-licensed and auditable end to end. Read exactly how your audio is handled, file an issue, or send a pull request. Standing on the shoulders of great open models and runtimes.
Free, 100% local AI voice dictation for Windows. Hold a key, speak, release — clean text at your cursor in any app.