Repository navigation
Releases
Kairo 0.1.3
Fix
- Long texts in the actions window. Anything over about 4,000 characters used to be cut into chunks and sent one after the other, so a whole spec meant several requests and Summarize quietly summarized each chunk before summarizing the summaries. Rewrites now go out in a single streamed request up to about 32,000 characters, Summarize up to 250,000. Only longer texts still fall back to chunks.
Kairo 0.1.2
Onboarding
On first launch Kairo now walks you through what it does instead of dropping you on an empty dashboard. Seven short screens, one sample answer: you grant the Accessibility permission, read the sample out loud with ⌥A, then summarize it, translate it and paste the result back with ⌥T. The AI steps use pre-written results, so no API key or license is needed to try them. Each step unlocks when you actually do it, and the reading step goes through the real player, so it ends up in your history like any other reading.
Skip is always available. Once finished or skipped, the flow never shows again. Existing installs won't see it either.
Fixes
- Copy-on-select terminals. In terminals that copy the selection as you drag and ignore Cmd+C (iTerm2 with the option on, Orca), ⌥A and ⌥T found nothing to read. Kairo now falls back to what those terminals already put on the clipboard.
- Emoji are no longer read aloud. Voices used to say "rocket" for 🚀. They're stripped before speaking, both for the macOS voice and the OpenAI ones.
Kairo 0.1.1
First public build of Kairo.
Kairo reads whatever text you have selected out loud, or rewrites it with AI, from a global shortcut. You never leave the app you're in. I made it because I spend most of my day reading long answers from Claude Code and ChatGPT and I'd rather listen to them.
Reading (free)
Select text anywhere, press ⌥A. A small player appears top right with the transcript, and words light up as they're read. ⌥P pauses, ⌥S cycles through five speeds (0.75× to 2×), Esc closes. Works over fullscreen apps. The player never steals focus.
The voice is the macOS one, picked from the language of the text. French works fine. The Home page keeps your last 50 readings if you want to replay one. Shortcuts can all be changed on the Shortcuts page.
Kairo Pro
Pro is a one time license, no subscription. It unlocks everything that goes through OpenAI, with your own API key. I don't take a cut on usage. OpenAI bills your key directly and a reading costs cents.
What it unlocks:
- Nine OpenAI voices. Audio is streamed, so it starts playing before the whole file is back.
- AI actions on ⌥T: Humanize, Translate (French, English, Spanish, German, Italian), Summarize. The prompts are tuned for developer text. Code names stay as they are, code blocks get described in one sentence, and AI filler and markdown get dropped.
- The result streams into a card. Listen to it, replace the original selection with it, copy it, or hit Then to chain another action. A breadcrumb takes you back to any step, with a before/after toggle.
- Long selections are chunked. Very long summaries take two passes.
- Text model and speech model are your choice in Settings. Each option shows a rough cost so you know what you're paying OpenAI for.
- A default output language for Humanize and Summarize, or leave it on auto to follow the text.
One license works on up to three Macs. Unlink a Mac from the Kairo Pro page to free the slot.
Updates
Kairo checks for new versions by itself and offers them in the sidebar. There's a manual check in Settings too.
Requirements
- macOS on Apple Silicon (Kairo_aarch64.dmg) or Intel (Kairo_x64.dmg). Signed and notarized, so no Gatekeeper warning.
- The Accessibility permission. That's how Kairo grabs the selection (it simulates a copy and restores your clipboard after). The app points you to System Settings the first time.
- An OpenAI API key, for Pro only.
It's a first release, so there will be rough edges. Open an issue if something breaks.