Follow the conversation.
A movable caption bubble turns system audio into readable words. Choose realtime Nemotron captions or a more relaxed, balanced pace.
A little clarity for everything you hear.
Read what you hear.
Keep what matters.
Live captions for your Mac’s audio. A second language when you need it. Dictation and transcripts when you want to keep the words.
macOS 26+ · Apple Silicon · No account or API key
Your audio stays here. Recognition and translation run on your Mac.
How privacy worksFrom listening to doing
The words you need.
The way you need them.
Keep both languages.
Put a local translation below the original. Translate complete sentences, or turn on realtime previews as the sentence grows.
Speak. Release. Done.
Hold your dictation shortcut and speak. The result goes into the field you started in when possible, and stays on the clipboard.
Take the words with you.
Save bilingual transcripts, search your sessions, or export SRT, WebVTT, Markdown and plain text. Add a local summary to revisit the essentials.
A native Mac app
Set it up once.
Make it yours.
Download the signed, notarized app. Move it to Applications. Setup walks you through the models and permissions for the features you choose.
Prefer a ZIP? Download it here ↗Choose captions, dictation, or both.
You decide what to enable. There’s no account to create.
Download your local models.
Weights are downloaded on first use, then stay on your Mac. Check the available storage in the app before downloading.
Grant only the access you need.
Captions use macOS Screen Recording permission for system audio. Dictation uses the microphone and Accessibility for insertion into other apps.
Private by design
On your machine.
On your terms.
Audio, recognition, translation and saved transcripts stay on the device. No audio uploads, account, analytics or API key.
Model downloads need an internet connection. Update checks are opt-in. The captions microphone mixer is off by default; dictation uses it only while you hold the shortcut.
A few useful details.
Will it work on my Mac?
Syllivra requires macOS 26 or later and Apple Silicon. Intel Macs and Windows are not supported releases. Model speed depends on your Mac, available memory and the models you choose.
Does it need the internet?
To download models and, if enabled, check for updates. Once the chosen models are installed, recognition and translation run locally.
Can I use more than one language?
Yes. Language coverage depends on the recognition and translation engines you select. Dictation without translation preserves supported spoken languages; turning translation on asks for the result in your chosen target language.
How do local summaries work?
Since version 2.6.6, Syllivra opens Summarize, Summarize Again and Remove Summary in the resizable transcript reader. It uses the Gemma cleanup model already installed for dictation and does not download another model. Summaries can miss or invent details, and long transcripts may omit earlier lines. Original spoken lines and subtitle exports are preserved.
Let the words stay with you.
Download Syllivra for Mac ↓Version 2.7.0 · Signed and notarized