Features
Record the room, the call, or both
Pick the microphone, the sound playing on your Mac, a single app, or any mix; one recording, one transcript, each line marked with where it came from. macOS asks once for microphone and once for system-audio permission.
Read it as it’s said
Captions appear while you record; a Translate-to control offers English, 中文 or Español; macOS downloads a language pack with its own dialog the first time. The saved transcript is a separate, full-quality pass and may differ from the live captions.
Who said what, in about 100 languages
A full transcript after the recording, in the language spoken, mixed languages included; speakers told apart; name a speaker and later recordings greet them by name with a confidence shown, undone with “Not them”. On macOS 26 with Apple Intelligence turned on, titles come from the recording’s own content, written on your Mac; otherwise a recording shows its date and time.
Everything you’ve recorded, searchable
Folders, favorites, notes anchored in a recording, and a Trash that keeps things 30 days; search by words or by meaning.
Drop in the documents too
Import audio and video (the audio is kept), transcripts (VTT, SRT, text), documents (PDF, Word, Markdown) and images, whose text is read on your Mac; all searchable with the recordings.
Your iCloud, your other Mac
Turn on iCloud sync and your transcripts, notes, folders and speaker names reach your other Macs through your own iCloud account. Recordings and documents can travel too, if you say yes on each Mac; a file downloads only when you open it. We run no server and receive nothing. Each Mac uploads its own recordings only after you say yes on that Mac; a file over 260 MB stays on its Mac and its transcript still syncs; erasing a recording removes it from iCloud too.
Your library, your assistant, your Mac
Aftersaid runs a Model Context Protocol server on your Mac for tools you run there; read tools for sessions, documents, speakers, search and folders, and folder writes behind a separate token; nothing listens off the machine.
The models are in the app
Speech recognition from OpenAI’s Whisper models converted by us and shipped inside the app, speaker separation and search embeddings on-device, titles from Apple’s on-device model; a fresh install transcribes with no network.
