Skip to content

Aris 1.2

Choose a tag to compare

@LumenJSN LumenJSN released this 14 Nov 19:52
· 1 commit to main since this release

Aris 1.2 brings talking and dictation to macOS 14 Sonoma and macOS 15 Sequoia. Until now Aris could only listen on
macOS 26; on older systems it can now hear you too.

What's new

  • Talking and dictation on macOS 14 and 15. On these systems Aris listens with Apple's older speech recognition
    engine. Holding the talk key, hands-free conversation (with barge-in), the wake word, spoken answers to cards and
    dictation into any app (with the pill) all work. macOS 26 keeps using Apple's newer engine, as before.
  • On the Mac where it can be. When your language has an on-device speech model, the older engine hears on the
    Mac. When it has none, what you say goes to Apple's servers to be recognised; Aris tells you once, on a card, the
    first time that happens. Aris hears the first of its languages the engine has (Korean by default).
  • Turn on macOS Dictation. On macOS 14 and 15, Apple's speech recognition only works while macOS Dictation is on:
    System Settings › Keyboard › Dictation. If it is off, Aris says so instead of hearing nothing: a talk turn ends
    with that line, and a dictation shows it in the pill.
  • Long dictations. The older engine stops after about a minute. Aris starts a fresh one at a pause just before
    that and stitches the text together in order, so a long dictation or a long turn carries on without a gap.
  • Which engine is listening. Settings › Voice · Language › Listening engine shows the engine in use, for
    example "Apple speech recognition on this Mac (macOS 15)" or "… on Apple's servers".
  • Polish needs macOS 26. Dictation on macOS 14 and 15 writes what you said, with your learned dictionary;
    polishing it in your own writing style uses Apple's on-device language model and needs macOS 26.

Fixes

  • The "talking needs macOS 26" card and the matching line in Settings are gone; the notch's glass look and dictation
    polish are still the only parts that need macOS 26.
  • If speech recognition stops right after it starts, three times in a row, Aris says "Speech recognition stopped right
    after it started. Please try again in a moment." instead of ending the turn with nothing heard. Words already
    written still count.

Requirements

  • A Mac with Apple silicon and macOS 14 Sonoma or later (the notch's glass look and dictation polish need
    macOS 26; older systems get a plain dark notch; Macs without a notch work too).
  • On macOS 14 and 15, macOS Dictation turned on for talking and dictation (System Settings › Keyboard ›
    Dictation).
  • At least one brain: Claude Code, Codex or Antigravity installed and logged in, or an API key for a model service
    (local models in Ollama or LM Studio work too).

Install or update

  1. Download Aris-1.2.0.dmg below (the onboarding's sounds and the recorded Korean greeting are inside the app).
    Aris-1.2.0.zip is the same app, for those who prefer a zip.
  2. Quit Aris if it is running. Open the DMG and drag Aris onto Applications (replace the old copy).
  3. Open Aris. The first time, macOS says it could not verify Aris: click Done.
  4. Go to System Settings › Privacy & Security, scroll down, click Open Anyway next to the Aris message and
    confirm. Aris opens and sets up its background helper.

With Homebrew instead: brew upgrade --cask aris (or, the first time, brew tap lumenjsn/tap,
brew trust lumenjsn/tap and brew install --cask aris). The cask clears the download quarantine flag, so step 4 is
not needed.

Why step 4: Aris is ad-hoc signed and not notarized by Apple; the project has no paid Developer ID. macOS
blocks a downloaded app like that until you allow it once. (From Terminal instead:
xattr -dr com.apple.quarantine /Applications/Aris.app.) Only do this for a copy downloaded from this page, and check
it first: shasum -a 256 -c SHA256SUMS in the download folder, or compare with the sums below.

Permissions after an update: each new download is a new program to macOS, so grant the permissions again. If a
switch in System Settings already shows on, turn it off and on, or remove the row with − and add Aris again. Your
settings, memories and saved steps are kept. Listening on macOS 14 and 15 uses the same Microphone and Speech
Recognition permissions as before; there is no new one.

Building from source works too (see the README); the onboarding sounds of this build are not in the repository, so a
build from source runs the onboarding without them and the Mac's own voice speaks the greeting.

Aris lets an AI read and act on your Mac. It asks before anything that writes, sends or deletes, but it is a personal
project without a warranty: read the README's safety model first.

SHA-256

f6609b877e117ff8b2a3d0f52773ec02a79e98160882614b8fbec267eae5edee  Aris-1.2.0.dmg
38957a50b21b1772f3f8fc26992a980143a6a02b8ab2292e773ba0e037b37db6  Aris-1.2.0.zip