CAPTUREScreen, GIF, scrolling capture or clipboard signal.
ANNOTATIONMark, number, box or arrow tied to a speech moment.
WORKFLOWRepeatable working flow.
This guide was produced with the very method it describes: one 12-minute spoken session, rectangles drawn while talking, and captures taken along the way. Every visual below is interactive — hover or click the numbered targets and the explanation opens next to the target.
Everything on this page was captured in a single 12-minute conversation; transcript, marks and visuals share one timeline.
The transcript view of the conversation that produced this guide; the content was narrated in a single voice session.
Start recording with ⌥A. The recording bar appears with the Capture Actions menu next to it; while you talk, everything is collected with timestamps.
While recording, every capture action lives in one menu, each with its shortcut.
The recording bar and the seven targets of the Capture Actions menu.
Capture a screen region when you need to show something visual; the capture lands in the conversation queue — hover for a preview, click to edit.
Hover: preview. Click: edit. The bar below shows which agent and voice profile you are talking to.
Preview of captured content inside the conversation, the edit area and agent info.
In the editor the rectangle tool comes pre-selected: say "this part needs fixing", draw a rectangle, and the pause matches the drawing to that exact moment of speech.
Draw rectangles, drop numbers if you prefer; remove a wrong mark by tapping it or with ⌘Z.
Editor toolbar: pre-selected rectangle tool, number tool and undo.
References embed into the flow of speech; a rectangle chip binds to the moment it was drawn, an image chip to the original file.
Type, source and drawing references embedded as chips in the transcript line.
Add Note attaches typed text to that moment of the conversation — use it sparingly, for proper nouns and words the transcriber gets wrong.
For special words only; Enter attaches the note.
The Add Note panel; typed text attaches to the speech timeline.
Use Get Text to pull text straight out of an image, and GIF Record to show anything that moves.
The selected region is recorded; Stop finishes, Cancel discards.
The panel shown while a GIF records: duration, size, stop and cancel.
You watch the GIF; the agent receives a frame digest of changed regions instead.
A real GIF captured during the session.
Only changed regions become frames — moving objects and changing text are captured one by one.
Changed-region frames produced from the GIF, with the Frames badge.
Watch the GIF to verify; switch what the agent receives from the selector.
GIF preview and the GIF/Frames format switch.
For time-based work, the Speech Context menu marks a start flag (press again to close the range) and Speech Source hands the agent the fully timestamped transcript — it finds the exact second you said "I clicked this and that happened".
Your speech is data; flags define ranges, the source carries exact time.
Speech Context menu: start marking and the timestamped speech source.
A long page becomes one tall image instead of two or three fragments.
Scrolling Capture under Special Capture.
It is all one flow: start with ⌥A, talk, mark what you see, capture motion as GIF, text as text, special words as notes. When the session ends, the agent holds your speech, marks and visuals merged on a single timeline.
