Desktop App v0.3.7 (superseded by v0.3.8)
Superseded by v0.3.8 on Windows.
There is no macOS build of 0.3.8 yet, because it has to be produced on a Mac,
so the macOS download on this page is still the current one. It does not
carry 0.3.8's fixes to the 3D avatar. Its 2D renderer is unaffected and is the
accurate one to use on a Mac for now.
A floating BSL signer that reads what your computer is saying, so anything it
plays can be signed, not just what is in a browser. Beta on both platforms.
Install on Windows
Download Signlytic AI Setup 0.3.7.exe below and run it. Windows SmartScreen
will warn that the publisher is unknown, because the build is not code-signed
yet: choose More info, then Run anyway.
Install on macOS
Download Signlytic AI-0.3.7-universal.dmg below. One file covers both Apple
silicon and Intel, which is why it is larger than the Windows installer.
macOS 13 or newer.
The build is neither signed nor notarised, and macOS is stricter about that than
Windows is. Opening it normally will fail. Instead, drag the app to Applications,
then right-click it and choose Open, and confirm once. After that it launches
normally.
The Mac app asks for permission the first time you press Start Captions:
microphone, speech recognition, and screen recording if you pick system audio.
Screen recording is how macOS gates capturing what other apps are playing.
macOS is newer and less tested than Windows
The Windows app has been through several releases. The Mac one is on its first,
so expect rougher edges. Microphone is the default and is the best tested path.
The Live Captions option under "Listen to" is experimental. Its ability to
read the Live Captions window was proven with a standalone probe, but it was
never seen delivering a sentence end to end inside the packaged app, because the
Accessibility permission it needs was never granted before that work stopped. If
it does nothing for you, that is the known gap: use microphone or system audio
instead. The source is kept on the macos-live-captions-unverified branch of the
main repository, with the remaining check written down.
What changed in 0.3.7
macOS support. Instead of reading system captions the way the Windows build
does, the Mac helper recognises speech directly, from the microphone or from
system audio, and feeds the same sentence assembler. Windows is unchanged by
this work and was re-verified after it.
The 3D avatar is about ten times smaller, 35 MB down to 3.5 MB for the male
model and 53 MB down to 5.4 MB for the female one. It appears in roughly half
the time, and far faster than that on a slow connection.
Only the textures were re-encoded. The skeleton, skinning and poses are the same
rig, so the signing itself is unchanged.
The app caches the model on disk, so this release also carries a version marker
that replaces the old copy rather than leaving it stranded.
Earlier releases
- 0.3.6 settings panel behind a gear, and microphone audio made optional
- 0.3.5 switches on "Include microphone audio" in Live Captions for you
- 0.3.4 signing speed: 1x through 2x, remembered between runs
- 0.3.3 fixed Live Captions in the installed build
Using Live Captions on Windows
Press Start Live Captions. The app starts Windows Live Captions for you if it
is not already running. On a machine that has never used it, Windows asks you to
accept a notice and download a speech model once; complete that and press again.
Live Captions needs Windows 11 22H2 or newer.
Attribution
Signlytic AI, by Oke Iyanuoluwa Enoch. If you use, fork or build on this project,
please credit the project and link back to this repository.