Transcripts

The Transcripts App turns any audio or video file into a searchable, navigable transcript with speaker detection. Transcripts are first-class entities — they live in your tree, they're searchable, and the AI can reason over them like any other content.

How transcripts come into being#

• Upload — drop in an audio or video file
• Video Studio — every ingested video produces a transcript automatically
• Calendar integration — meeting recordings can transcribe automatically when they finish
• Browser Extension — capture YouTube videos with their transcripts
• Computer Use — record a meeting via Computer Use, then transcribe

What the transcript gives you#

• Speaker-labeled lines — automatic detection (configurable)
• Timestamps — click any line to jump to that moment in the source
• Search — full-text across every transcript
• Cross-link — transcripts link back to the source media entity and forward to anything the AI extracts

AI patterns over transcripts#

• "Summarize this meeting" — a tight summary with speaker attribution.
• "What did Maya commit to?" — pulls Maya's action items from the transcript.
• Action item extraction — turns the transcript into Tasks linked to the meeting.
• Standup synthesizer — runs after every team standup; reads the transcript and updates the project doc.

Speaker management#

After the first run, the AI guesses who's who based on patterns. You can label speakers explicitly — once you've named "Maya" and "Diego" once, future transcripts use those names automatically.

Tips#

• Long recordings produce long transcripts. For meetings >1hr, ask for chapter summaries instead of one big summary.
• Hook a Transcript Synthesizer Agent to transcript.ready events — every meeting becomes structured action items the moment it finishes processing.
• For privacy-sensitive content, run transcription via a local Whisper model on the desktop app.