Get up and running in 5 minutes
Everything you need to install AddSub, choose a language pair, and start reading live subtitles.
1 · Download & install
Download the .dmg and drag AddSub into Applications. On first launch, macOS may ask to allow AddSub to capture audio — this is required and used only locally.
2 · Install the audio loopback driver
AddSub routes your Mac's audio through a small system driver so it can read any app's output. The driver ships inside the app — on first launch (or from Settings → Audio & System → AddSub Loopback Driver → Install Driver) AddSub presents a one-time admin password prompt to install it. Your audio is processed entirely on-device and never leaves your Mac.
3 · Pick your language pair
In Settings → Languages, choose the spoken language (or Auto-Detect) and your preferred translation target — e.g. Japanese → English. The overlay shows both original and translation.
4 · Choose a model
AddSub ships with on-device speech models. Start with Tiny (16 MB, fastest) and upgrade to Small or Medium for higher accuracy. Large models download once and are cached.
5 · Play anything & read captions
Start any video, stream, or call. AddSub transcribes and translates in real time, floats subtitles above the content, and labels speakers (S1, S2, S3…) as they talk.
Troubleshooting
No captions appear
Check Settings → Audio & System that the loopback driver is installed and routing is active. Confirm the app has Microphone permission in System Settings → Privacy & Security → Microphone. Playback must be on this Mac's output.
Why does macOS show 'Speech data will be sent to Apple'?
This is Apple's mandatory, hardcoded system disclaimer shown whenever an app requests macOS speech recognition permissions. AddSub processes 100% of speech locally on your Mac's Neural Engine — zero audio or transcripts are sent to the cloud. You can also switch to the standalone Whisper.cpp engine in Settings → Models, which requires zero speech permissions.
Apple Speech vs Whisper.cpp engine
AddSub includes two native engines: Apple SpeechTranscriber (ultra-fast, instant live streaming on macOS 26) and Whisper.cpp (100% self-contained offline neural models). Choose your preferred engine in Settings → Models.
Captions lag behind audio
On Apple Silicon, use a faster model (Tiny or Small). Ensure no other heavy Neural Engine workload is running. Latency is typically under a second.
Translation looks odd
Some languages translate more naturally one way. Try switching the pair (e.g. EN → JA vs JA → EN) or using Auto-Detect for the source.
Speaker labels are wrong
Diarization improves over the first ~30 seconds as it clusters voices. For best results, keep the microphone channel free of background noise.
Driver install fails
Run 'AddSub' from Applications after installation, not from the mounted DMG. If it still fails, contact support — we'll get you set up.
How do I reset the trial?
The 7-day trial is tied to this Mac and starts at first launch. Reinstalling won't extend it. After it ends, activate a one-time license in Settings → License.