Press record. Read the notes.
What is recorded.
Your microphone, your Mac’s sound or both. System sound is a switch, off until you turn it on, and needs macOS’s Screen & System Audio Recording permission. Wear headphones when you record both. You can also import an audio or video file.
Recent
- Design review
- Onboarding interviews, round 2
All notes
Your conversations, all in one place.
Where it lives.
Audio is saved as short files in a folder on your Mac, Library/Application Support/Trilsy/Recordings. The transcript and notes are files next to it. Trilsy has no account and uploads nothing. Your Mac’s disk encryption protects them; Trilsy does not add its own.
How the transcript is made.
After you stop and press Create notes, Trilsy transcribes on your Mac with Whisper (the small model). Language is detected, or you pick one. You can add words it should know.
How the notes are written.
A language model on your Mac writes detailed notes section by section, then puts a short overview on top. Check them against the audio; the model can be wrong, and a timestamp it writes into the notes is not a link.
What you can do with them.
Edit the notes, the transcript and the title. Search your recordings by title, transcript and notes. Export Markdown (notes and transcript), text (transcript) or SRT subtitles (transcript).
What you need.
Apple silicon, macOS 15 or later, 16 GB of memory recommended. The two models download once from Hugging Face and take several GB; recordings are saved uncompressed, so they are large.
The code.
The source is on GitHub.
View on GitHub