Technique / CaptionCut

Keep a caption outside the shot transform

Retain identity while the underlying shot changes scale.

Run the original

4.8 s · 30 fps · silent

Original neutral assets · 4.8 seconds · phase values are chosen defaults, not source timings.

npm run render -- CaptionCut out/CaptionCut.mp4 --codec=h264 --muted

Download the complete pinned project · Inputs, phase meanings & full source code

Official source / inspected interval

ChatGPT can now complete tasks on your computer

Watch the official OpenAI upload ↗ · Published 2026-07-16T11:00:16-07:00

Inspected 6.15–12.8 s. Consecutive subset: 9.8–10.4 s, 15 frames. Dates come from the official watch-page publication field.

Last wide shot; identifier at lower left at source 9.884875 seconds — 9.884875 s · Last wide shot; identifier at lower left

Closer shot; identifier stays screen-anchored at source 9.926583 seconds — 9.926583 s · Closer shot; identifier stays screen-anchored

Caption holds independently of subject motion at source 10.135125 seconds — 10.135125 s · Caption holds independently of subject motion

Close shot still carries the identifier at source 11.594917 seconds — 11.594917 s · Close shot still carries the identifier

Source: OpenAI. Bounded excerpt and annotated stills for the analysis below; not included in the project download.

What changes across the edit

Shot boundary: Source frame 237 at 9.884875 s is the wider presenter shot; frame 238 at 9.926583 s is a closer angle.
Framing: The person changes scale and background framing. Name and role remain at the same screen location.
Layer ownership: Live-action shot beneath independently positioned white identifier text.
Motion: One direct shot replacement; caption does not scale with the person.
Typography: Name and role remain on one horizontal line. Native font and backing treatment are not recovered.
Action order: Presenter → identifier appears → cut to closer presenter with identifier retained.
Pacing & readable holds: Identifier survives the 9.926583 s cut and remains through the sampled close shot; it disappears before the later screen-demo composition.

Recreation versus observation

Matches

Caption persists unchanged in screen space while the original workspace below it cuts to detail.

Differs

An artifact label on a contrast panel, not a speaker lower third or live-action recreation. It does not combine a talking head with presentation video.

Measured

Adjacent cut frames 237 / 238 at 9.884875 / 9.926583 s.

Inferred

Screen anchoring is visible; actual compositing software is unknown.

Chosen defaults

All Remotion phase frames, cubic Bézier (.16,1,.3,1), geometry, colors, font choice and reading holds are chosen defaults. None is a recovered OpenAI specification.

{
  "title": "Make room for the idea.",
  "prompt": "Arrange a small exhibition.",
  "result": "Three studies, ready to compare.",
  "label": "EXHIBITION / 03",
  "accent": "#40685b",
  "action": 18,
  "change": 72,
  "settle": 72,
  "hold": 72,
  "focusX": 0.66,
  "focusY": 0.64,
  "zoom": 1.55
}

Increase hold to extend only the final quiet state. Change the phase frames to change choreography; preserve their ordering. Content limits and the exact meaning of each phase are in the handoff.

Exact original geometry, layer ownership and asset contracts · Schedule against an actual recording