Screen Record captures your screen straight onto the timeline — and, because it tracks where you click while recording, it automatically builds a smart-zoom camera move that pushes in on each click and eases back out. The result feels like a hand-edited product demo with no manual keyframing. Optional AI understanding toggles go further: they transcribe the take, generate chapters and a summary, lay a storyboard on the board, and let you ask the board agent questions about the recording.
Start a recording
Click Screen Record in the timeline toolbar. A floating Screen Recording setup panel opens (it stays on top, so you can move it anywhere on screen):
- Capture — Display (full screen), Window (you pick the window each time), or Area (draw a region, optionally saved as a reusable preset).
- Audio & camera — toggle System audio (on by default) and Microphone (off by default). With both on, the mic is mixed into the system audio.
- Hide app while recording — full-screen capture only. The app window hides the moment recording starts and comes back when you stop, so it never shows up in the take.
- AI understanding — three optional toggles, covered in the sections below: AI transcript & understanding, Auto chapters & summary (requires the transcript toggle), and Create storyboard.
The panel remembers your choices for next time. Click Start recording; the toolbar button turns into a red Stop button with an elapsed-time counter — click it (or use the OS “Stop sharing” control) to finish.
On macOS you’ll see a hint: “Enable Input Monitoring for this app so your clicks drive auto-zoom.” Granting Input Monitoring permission is what lets the recorder see your clicks and build the zoom layer.
What lands on the timeline
When you stop, the recording is uploaded and added as a video clip on the active video channel (a Screen recording channel is created if needed). If the capture is larger than the current project canvas, the project resizes up to fit it. The take is also dropped onto the board as a video card.
Alongside the clip, the recorder stores the cursor path and click events so it can drive auto-zoom.
Smart zoom (auto-zoom)
If clicks were captured, an Auto-zoom clip is added to the project’s camera channel — a keyframed virtual camera that pans and zooms over the whole stack. (Before 7.3 this was an adjustment layer that moved only the footage under it; as a camera move, anything you composite over the recording — callouts, cursors, logos — travels with the zoom instead of sliding out of frame.)
Clicks that happen close together in time and space are grouped into focus moments. For each one the camera pushes in (~2×), pans to centre the click, holds briefly, then eases back out — starting just before the click so the zoom is already settling as you click.
Record again and the new move joins the same camera channel after the existing ones — there is only ever one camera channel in a project.
The status line after recording tells you what happened:
- “Recording added · auto-zoom generated from your clicks”
- “Recording added · no clicks detected for auto-zoom”
- “Recording added · enable Input Monitoring for auto-zoom” (macOS permission not granted)
Adjust or remove the zoom
The auto-zoom is an ordinary camera clip, so every move is editable. Select it and the right panel becomes the Camera panel: swap in a motion preset, or use the keyframe editor to tweak Zoom, Pan X / Pan Y and easing on any key, add or delete focus moments, drag its edges to retime, or delete the clip to drop auto-zoom entirely.
A layer that should stay frame-locked through the zoom (a watermark, a fixed caption) can opt out with the camera button on its channel header.
AI transcript & understanding
Turn on AI transcript & understanding in the setup panel to transcribe the take in the background after you stop (a notice reads “Recording transcript ready” — usually within a minute or two) and make it agent-readable.
Once the transcript is ready, select the recording clip and the right panel shows an AI CAPTURED DATA section:
- Extract voice with transcript — lifts the spoken audio onto a voice channel (see Voice Channels and Transcribe Editor) as an already-transcribed voice take and silences the video, so you can edit the narration word-by-word.
- Add captions… — builds a timed caption track from the transcript. Pick a caption animation preset (or No Animation); captions follow the clip’s current trim.
The take also becomes something the board agent can reason about: ask it what happened in a recording, what was on screen at a given moment, or what you clicked — it reads the transcript, the click/cursor event track, and individual video frames on demand. See Agent-Driven Editing.
Auto chapters & summary
Requires AI transcript & understanding. After transcription, an AI pass detects scene changes, then titles and summarizes the take:
- The clip is renamed to the generated title, and chapter markers appear along the clip on the timeline.
- The recording gets a summary and tags the agent can use when you ask about your footage.
- A notice reads “Recording chapters & summary ready” when it lands (this runs after transcription, so give it a few minutes on long takes).
If the pass fails, ask the board agent to regenerate insights for the recording — you can also ask it to retitle the take, edit the chapters, or re-tag it.
Create storyboard
Turn on Create storyboard to lay the take’s detected scenes onto the board as a “Storyboard for …” scaffold — one frame per scene with timecode notes and a ready-made agent prompt (see Scene Frames and Storyboard). It’s placed below your existing board content, and scene detection runs in the background so long takes don’t hold up the editor.
Leave a Reply