# Clear a four-word title right half first, left half two frames later, then fade in a two-part lower third

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 can now complete tasks on your computer" (dB6pOolO7io), published 2026-07-16.
Source interval [128, 200) on the 24000/1001 fps master: 72 frames, 3.003000 s. Local frame 0 = source frame 128.
Create a 1920x1080 progressive timeline at 24000/1001 fps and a Fusion composition exactly 72 frames long. Keep integer frames; do not round the frame rate.
Background field: #d6d9d5. Export silent H.264 yuv420p to match the intended playback container.

## Observation
The four-word staircase title and its bottom-left sub-line sit over a presenter at a desk. At ~f15 the right-hand pair of words and the sub-line vanish together; two frames later (~f17) the left-hand pair vanishes, leaving clean footage. After an 18-frame pause a small lower third fades in at the bottom-left (~f35–f38): the name on the left and, with a gap, the role to its right on the same baseline. It holds to the end.

## Mechanism
Two-beat title clear (right group, then left group) followed by a single-baseline name-and-role lower third fading in after a pause.

## Phases (local frames)
- 0: Title and sub-line hold
- 15: Right words and sub-line off
- 17: Left words off
- 35: Lower third fades in
- 38: Hold

## Layer inventory (absolute 1920x1080 top-left pixels, back to front)
- bg [rect] at (0, 0) size 1920x1080 fill #d6d9d5
- window-a [rect] at (560, 0) size 420x760 fill #eef1ee
- window-b [rect] at (1040, 0) size 420x760 fill #eef1ee
- mullion [rect] at (760, 0) size 14x760 fill #2a2d2a
- mullion-b [rect] at (1240, 0) size 14x760 fill #2a2d2a
- desk [rect] at (0, 860) size 1920x220 fill #c9b89a
- shelf [rect] at (1500, 540) size 380x320 fill #8b9a9a
- figure [ellipse] at (760, 180) size 320x900 fill #1b1b1d
- laptop [rect] at (700, 760) size 440x110 radius 8 fill #2a2a2c
- w1 [text] at (72, 44) text "Desktop" 84px weight 600 #ffffff visible [0, 17)
- w2 [text] at (190, 140) text "and" 84px weight 600 #ffffff visible [0, 17)
- w3 [text] at (1180, 520) text "Browser" 84px weight 600 #ffffff visible [0, 15)
- w4 [text] at (1300, 620) text "Agent" 84px weight 600 #ffffff visible [0, 15)
- sub [text] at (120, 920) text "in the new companion app" 34px weight 500 #ffffff visible [0, 15)
- name [text] at (120, 990) text "Jonas Feld" 24px weight 600 #ffffff
    keyframes opacity: f35=0, f38=1 (out)
- role [text] at (470, 990) text "Developer Relations at Northwind" 24px weight 400 #ffffff
    keyframes opacity: f35=0, f38=1 (out)

## Construction steps
1. Timeline 1920x1080 at 23.976 fps; Fusion composition 72 frames. Local 0 = master frame 128.
2. Title: reuse the five Text+ nodes from the type-on study (84px semibold, sub-line 34px) at the same positions, fully revealed from frame 0.
3. Clear: set the clip end of word 3, word 4 and the sub-line to frame 15 (exclusive) and of word 1 and word 2 to frame 17. Hard cuts, no fades.
4. Lower third: two Text+ nodes on one baseline at top y=990, 24px; name at x=120 semibold, role at x=470 regular. Keyframe Blend 0→1 over frames 35–38 on both.
5. Hold the lower third with no motion until the end of the shot.
6. Easy mistake: clearing all four words at once. The right-hand group leaves two frames before the left-hand group.

## Content variants
Default copy:
- w1: Desktop
- w2: and
- w3: Browser
- w4: Agent
- sub: in the new companion app
- name: Jonas Feld
- role: Developer Relations at Northwind
Adaptation changes:
- w1: Inbox
- w2: and
- w3: Calendar
- w4: Sync
- sub: inside the assistant for Mac
- name: Ana Kovac
- role: Platform Engineering at Halcyon

## 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.
