Desktop app

OpenASR Desktop is a native macOS app — a closed-source shell (onboarding, UI, packaging, signing, auto-update) built over the same Apache-2.0 open core as the CLI. It adds no hidden transcription path: everything that touches your audio is the open code at github.com/QuintinShaw/openasr. Grab it from the download page.

First launch

The first launch walks through a short onboarding: it asks for Microphone access, then (on macOS) Accessibility access, then offers a quick practice run of the dictation shortcut. You can skip onboarding at any point — the app still opens, and each permission is requested again the first time a feature that needs it is used.

Permissions, and why each one exists

OpenASR Desktop asks for three macOS permissions. It does not use or request Input Monitoring — the dictation hotkey is built on NSEvent monitors that only need Accessibility, not the lower-level Input Monitoring TCC grant that a keyboard event tap would require.

Microphone

Needed for file-independent capture: live captions from your mic and the dictation shortcut both record through the microphone. The app requests it during onboarding, or the first time you start a mic-based feature if you skipped that step. Grant or revoke it in System Settings > Privacy & Security > Microphone.

Accessibility

Needed for dictation specifically: inserting recognized text into whatever app you’re focused in (and recognizing the global press-and-hold shortcut anywhere in the OS) both work through the macOS Accessibility APIs. Without this permission the dictation shortcut can still record, but it has no way to type the result into another app.

Grant it in System Settings > Privacy & Security > Accessibility — find OpenASR Desktop in the list and toggle it on (use the + button to add it if it’s missing). If you grant it while the app is running, switch back to OpenASR Desktop and it re-checks automatically within a couple of seconds — no restart needed.

System Audio capture

Needed only for live captions from system audio (captioning a call or a video instead of your own voice) — Microphone-only live captions never trigger this prompt. macOS asks the first time you click “Start system audio captions,” using Core Audio process taps (macOS 14.2 or later). Manage it in System Settings > Privacy & Security, under the system audio / screen recording section for OpenASR Desktop.

Choosing and downloading a model

The Models screen lists the signed catalog pulled from the same model-registry/catalog.json the CLI uses. Onboarding and first use offer a default model automatically; picking a different one, or a different quantization, is a click away on the Models screen. Every download is a verified .oasr pack — nothing installs silently, and nothing you haven’t asked for reaches the network. See Models for the catalog families, and Troubleshooting for what to expect size- and time-wise on a first pull.

Dictation and live captions

Dictation is a press-and-hold (or toggle, depending on your configured shortcut) that transcribes while held and inserts the result into whichever app is focused when you release it — this is what needs Accessibility.

Live captions streams a running transcript on-device as you speak, from either the microphone or system audio, in a small floating overlay you can drag anywhere on screen. Nothing here leaves the machine unless you explicitly opt into remote compute in Advanced Settings.

Auto-update

The app checks for updates against dl.openasr.org shortly after launch, and again whenever you click “Check for updates” in Settings. A check never installs anything by itself — downloading and installing a newer build, and relaunching into it, are separate steps you trigger yourself. Settings also lets you switch between the stable and preview release channels.

Uninstalling

Quit OpenASR Desktop.
Delete /Applications/OpenASR Desktop.app.
Optional — remove local data and settings: ~/.openasr/ holds your config, installed models, and (if you set one) your Hugging Face token. Deleting it removes downloaded models too, so only do this if you want a clean slate.
Optional — if you ever turned on “launch at login,” also remove the dev.openasr.desktop entry under System Settings > General > Login Items, or the ~/Library/LaunchAgents/dev.openasr.desktop.plist file if it’s still there.

Granted permissions (Microphone, Accessibility, System Audio) stay recorded against the app’s signed identity — you don’t need to reset them separately for an uninstall/reinstall to keep working, since a reinstalled signed build carries the same identity.

See Troubleshooting if dictation, the microphone, or the app itself isn’t behaving as expected.