Meet, Zoom, Teams, and any Mac audio - v1.0.50
Complete meeting recaps without a bot in the room.
Transcript records mic and system audio from any call, transcribes locally with Whisper, then hands structured context to Codex, Claude Code, Gemini, or OpenAI using your own setup.
Requires macOS 15.7 or later. No meeting bot, extension, or platform-specific recorder required.
AI connectors
Use the agents you already trust.
Transcript does not force a single recap backend. It prepares the transcript, calendar context, and speaker hints, then lets your Codex, Claude Code, Gemini, or OpenAI workflow produce the final notes.
Codex
Recommended for complete meeting recaps with strong structure and follow-up extraction.
Claude Code
Use Claude.ai subscription auth for long-form synthesis and nuanced written notes.
Gemini
Bring a Gemini key when you want direct model calls or audio-first Gemini processing.
OpenAI
Use OpenAI models directly when you want API-key based recap generation.
Why it matters
Capture anywhere. Keep the meeting private.
Any call
Meet, Zoom, Teams, browser audio, local video, or system sound.
No bot
Nothing joins the meeting, announces itself, or needs host approval.
Your AI
Codex, Claude Code, Gemini, or OpenAI run through your own setup.
Works with any meeting app
Google Meet, Zoom, Teams, Slack huddles, browser calls, screen shares, or any audio your Mac can hear.
No bot joins the room
Transcript records from your Mac, so there is no extra meeting participant, calendar guest, or bot account.
Deep recaps with your agents
Send structured meeting context to Codex or Claude Code using the accounts and tools you already use.
Local Whisper transcript
Transcribe privately on-device with Whisper before any recap provider sees the text input.
Speaker-aware recap input
Optional Local, Remote, and Overlap hints help recaps preserve who said what without saving duplicate audio.
Calendar context
Read-only Google Calendar matching adds title, attendees, and local-speaker clues when available.
Recap pipeline
Small cues, richer recaps.
Recap input is kept as structured WebVTT, but cues are split by speaker label, duration, and size. Codex gets tighter context for summaries, decisions, objections, and owner-specific follow-ups without changing the visible transcript.
00:00:06.640 --> 00:00:15.900 Local: Let's validate the recording flow. 00:00:16.000 --> 00:00:19.200 Remote: The demo audio is now playing. 00:00:19.200 --> 00:00:20.300 Local: Interesting, actually.
01
Capture
Record mic plus system audio from whichever meeting app you use.
02
Transcribe
Whisper writes a timestamped local transcript first.
03
Recap
Codex or Claude Code gets compact, speaker-aware meeting context.
04
Organize
Calendar metadata, tags, exports, and AI rename keep sessions findable.
No duplicate diarization audio.
The Local/Remote beta stores derived RMS and peak activity per second in a hidden metadata file. It is not speech and cannot be used to reconstruct the meeting.
Terminal-friendly.
Use the app, or automate the same flow with the bundled CLI.
transcript record --duration 60m
transcript transcribe latest
transcript recap latest