# Tick a working timer through tool steps, push the camera into the window, and pop a preview card

Manual DaVinci Resolve / Fusion construction guide generated from the recipe. It is not a tested .drp.
The accompanying Remotion project (recipe.json) is the executable definition; the numbers below are copied from it.
Neither construction establishes source fidelity: this is an original vector approximation of an inspected editing moment.

## Source and clock
Reference: OpenAI's "ChatGPT Computer Use: Now faster with Live Picture-in-Picture" (IYYTI73J7Rg), published 2026-07-10.
Source interval [440, 512) on the 24/1 fps master: 72 frames, 3.000000 s. Local frame 0 = source frame 440.
Create a 1920x1080 progressive timeline at 24/1 fps and a Fusion composition exactly 72 frames long. Keep integer frames; do not round the frame rate.
Background field: #071120. Export silent H.264 yuv420p to match the intended playback container.

## Observation
A white assistant window floats over a night sky. A user prompt bubble sits top-right; under it a grey 'Working for 4s' status, an explanatory paragraph and a tool step line. The timer ticks to 6s (~f10) and 7s (~f20) while the step line changes to 'Loaded tool' plus a second step (~f20). From ~f22 the camera pushes into the window (about 1.7x by ~f34); during the push a small rounded preview card with a slide thumbnail pops in at the right of the paragraph (~f30-36). The magnified view holds to the end.

## Mechanism
Status text ticks in discrete steps while an eased camera push-in magnifies the window and a preview card scales in mid-push.

## Phases (local frames)
- 0: Window at 1x, Working for 4s
- 10: Timer ticks to 6s
- 20: Timer 7s; steps update
- 22: Camera push-in starts
- 30: Preview card pops
- 34: Push settles at 1.7x

## Layer inventory (absolute 1920x1080 top-left pixels, back to front)
- stars1 [ellipse] at (300, 120) size 4x4 fill #ffffff
- stars2 [ellipse] at (1700, 200) size 3x3 fill #ffffff
- stars3 [ellipse] at (250, 800) size 3x3 fill #ffffff
- water [rect] at (0, 860) size 1920x220 fill #0b1e33
- window [group] at (400, 40) size 1120x980 radius 20 fill #ffffff
  - wd1 [ellipse] at (420, 58) size 13x13 fill #f25f58
  - wd2 [ellipse] at (440, 58) size 13x13 fill #fbbe3c
  - wd3 [ellipse] at (460, 58) size 13x13 fill #58c951
  - tab [text] at (400, 55) size 1120x0 text "Add slides from notes" 14px weight 400 #555555
  - attach-card [rect] at (1280, 110) size 200x130 radius 8 fill #f6f6f4 stroke #e3e3df 1px
  - attach-l1 [rect] at (1300, 132) size 120x6 radius 3 fill #cfcfca
  - attach-l2 [rect] at (1300, 148) size 160x6 radius 3 fill #cfcfca
  - attach-l3 [rect] at (1300, 164) size 140x6 radius 3 fill #cfcfca
  - attach-l4 [rect] at (1300, 190) size 150x6 radius 3 fill #f7c948
  - attach-t [text] at (1280, 248) size 200x0 text "Harvest / rough notes" 13px weight 400 #777777
  - bubble [rect] at (820, 280) size 660x76 radius 16 fill #f1f1ef
  - bubble-icon [rect] at (840, 302) size 22x22 radius 5 fill #2f7cf6
  - prompt [text] at (872, 296) size 590x0 text "Desk  add a few new slides to my Harvest keynote with the rough ideas from my notes" 17px weight 400 #1c1c1c
  - status1 [text] at (440, 390) text "Working for 4s" 16px weight 400 #8a8a8a visible [0, 10)
  - status2 [text] at (440, 390) text "Working for 6s" 16px weight 400 #8a8a8a visible [10, 20)
  - status3 [text] at (440, 390) text "Working for 7s" 16px weight 400 #8a8a8a visible [20, 72)
  - para [text] at (440, 432) text "I'm using the desk skill to operate Notes and the slide app directly. I'll add distinct slides via the / 'Add slide' button, using the two note sections as separate slide concepts." 17px weight 400 #2b2b2b
  - step-icon1 [rect] at (440, 506) size 14x14 radius 3 stroke #9a9a9a 1.5px
  - step1 [text] at (464, 502) text "Reading desk skill" 15px weight 400 #8a8a8a visible [0, 20)
  - step2 [text] at (464, 502) text "Loaded tool" 15px weight 400 #8a8a8a visible [20, 72)
  - step-icon2 [rect] at (440, 538) size 14x14 radius 3 stroke #9a9a9a 1.5px visible [20, 72)
  - step3 [text] at (464, 534) text "Open Harvest keynote" 15px weight 400 #8a8a8a visible [20, 28)
  - step3b [text] at (464, 534) text "Thinking" 15px weight 400 #8a8a8a visible [28, 72)
  - composer [rect] at (440, 880) size 1040x100 radius 18 fill #ffffff stroke #e3e3df 1.5px
  - composer-t [text] at (462, 900) text "Ask for follow-up changes" 16px weight 400 #9a9a9a
  - approve [text] at (462, 946) text "+   ⟳ Approve for me  ⌄" 14px weight 400 #2f7cf6
  - model [text] at (1300, 946) text "5.6  Sol   Medium ⌄" 14px weight 400 #555555
  - send [ellipse] at (1436, 936) size 30x30 fill #111111
  - send-sq [rect] at (1446, 946) size 10x10 radius 2 fill #ffffff
  - preview [group] at (1100, 480) size 300x200 radius 14 fill #ffffff visible [30, 72)
      keyframes scale: f30=0.6, f36=1 (out)
      keyframes opacity: f30=0, f34=1
    - pv-bg [rect] at (1100, 480) size 300x200 radius 14 fill #eef1f5
    - pv-sidebar [rect] at (1112, 492) size 54x176 radius 6 fill #dfe4ea
    - pv-th1 [rect] at (1118, 500) size 42x26 radius 3 fill #ffffff
    - pv-th2 [rect] at (1118, 532) size 42x26 radius 3 fill #ffffff
    - pv-th3 [rect] at (1118, 564) size 42x26 radius 3 fill #ffffff
    - pv-slide [rect] at (1178, 502) size 208x156 radius 6 fill #f6f1e6
    - pv-title [text] at (1190, 512) size 184x0 text "Regional traditions take root" 13px weight 700 #4a3a2a
    - pv-l1 [rect] at (1190, 554) size 90x5 radius 2 fill #c9bfae
    - pv-l2 [rect] at (1190, 566) size 80x5 radius 2 fill #c9bfae
    - pv-l3 [rect] at (1190, 578) size 86x5 radius 2 fill #c9bfae
    - pv-art [ellipse] at (1296, 556) size 78x90 fill #b98a5a

## Camera (whole scene, about canvas centre)
- scale: f22=1, f34=1.7 (inOut)
- tx: f22=0, f34=136 (inOut)
- ty: f22=0, f34=119 (inOut)

## Construction steps
1. Timeline 1920x1080 at 24 fps; Fusion composition 72 frames. Local 0 = master frame 440.
2. Night background #071120 with a #0b1e33 water band and a few star dots; window plate 1120x980 white at 400,40, radius 20, heavy shadow.
3. Status: three Text+ nodes at 40,350 (window space) swapped with hold keys at frames 10 and 20 ('Working for 4s' -> '6s' -> '7s'). Step line 1 swaps to 'Loaded tool' at 20 and a second step appears; the second step becomes 'Thinking' at 28.
4. Camera Transform (centre pivot): Size 1 -> 1.7 and Center offset (0,0) -> (136,119) px over frames 22-34 with ease in/out, then hold.
5. Preview card: 300x200 white plate at 700,440 (window space) with a mini sidebar and slide; Size 0.6 -> 1 over frames 30-36 ease-out and Blend 0 -> 1 over 30-34.
6. Easy mistake: popping the card before or after the push. It lands inside the camera move (frames 30-36 of a 22-34 push), so the two motions overlap.

## Content variants
Default copy:
- tab: Add slides from notes
- prompt: Desk  add a few new slides to my Harvest keynote with the rough ideas from my notes
- attach: Harvest / rough notes
- status1: Working for 4s
- status2: Working for 6s
- status3: Working for 7s
- para: I'm using the desk skill to operate Notes and the slide app directly. I'll add distinct slides via the / 'Add slide' button, using the two note sections as separate slide concepts.
- step1: Reading desk skill
- step2: Loaded tool
- step3: Open Harvest keynote
- composer: Ask for follow-up changes
- slideTitle: Regional traditions take root
Adaptation changes:
- tab: Summarise the survey
- prompt: Desk  turn the survey export into a three-chart summary in my spreadsheet
- attach: Survey / export.csv
- para: I'm using the desk skill to operate the spreadsheet directly. I'll build three charts from the / response columns and place them on a new summary sheet.
- step1: Reading desk skill
- step2: Loaded tool
- step3: Open survey workbook
- slideTitle: Response summary

## Limits
Easing names map to Fusion spline shapes: linear = linear, in/out/inOut = ease curves on the same frames, hold = step, spring/overshoot = slight overshoot past the target before settling.
Text reveal frames give the first and last unit; distribute characters/words evenly between them. Cursor click rings last 14 frames.
Vector illustration is an approximation of layout and motion, not the product UI or recovered production assets.
