# Cut between speakers: the outgoing lower third drops with the picture and the new one rebuilds name-then-role

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 "Previewing Ultrafast mode: GPT‑5.6 Sol at up to 14X the speed" (EDWbX_-RYws), published 2026-08-14.
Source interval [1636, 1726) on the 24000/1001 fps master: 90 frames, 3.753750 s. Local frame 0 = source frame 1636.
Create a 1920x1080 progressive timeline at 24000/1001 fps and a Fusion composition exactly 90 frames long. Keep integer frames; do not round the frame rate.
Background field: #2a2420. Export silent H.264 yuv420p to match the intended playback container.

## Observation
Local 0-2 (master 1636-1638) show speaker A against a bright room with a complete lower third (bold name, then role to its right) at bottom-left. At local 3 (1639) a hard cut lands on speaker B in a warm beige room framed by dark pillars, and the lower third is gone. At local 6 (1642) speaker B's name pops on in the same position; at local 10 (1646) the role pops on to its right. Both hold to the end of the window with no motion.

## Mechanism
Lower-third continuity across a speaker cut: the graphic is cut out with the outgoing shot and rebuilt in two hard steps (name, then role) a few frames into the new shot.

## Phases (local frames)
- 0: Speaker A with full lower third
- 3: Hard cut to speaker B, graphic gone
- 6: New name pops on
- 10: New role pops on
- 89: Hold

## Layer inventory (absolute 1920x1080 top-left pixels, back to front)
- shot-a [group] at (0, 0) size 1920x1080 visible [0, 3)
  - a-wall [rect] at (0, 0) size 1920x1080 fill #e6e9ec
  - a-window [rect] at (60, 0) size 420x1080 fill #bfe0f0
  - a-door [rect] at (700, 0) size 600x1080 fill #f4f4f2
  - a-body [rect] at (560, 700) size 800x600 radius 220 fill #3a3a3a
  - a-head [ellipse] at (810, 220) size 300x400 fill #b47f5a
  - a-hair [rect] at (800, 190) size 320x130 radius 60 fill #1b1512
  - shot-a-floor [rect] at (0, 880) size 1920x200 fill #0e0b09
  - a-name [text] at (140, 954) text "Sayid Okonkwo" 28px weight 700 #ffffff
  - a-role [text] at (500, 954) text "Member of the Platform Team, Northwind Labs" 28px weight 500 #ffffff
- shot-b [group] at (0, 0) size 1920x1080 visible [3, 90)
  - b-wall [rect] at (0, 0) size 1920x1080 fill #a08a6c
  - b-pillar-l [rect] at (0, 0) size 110x1080 fill #1a1614
  - b-pillar-r [rect] at (1780, 0) size 140x1080 fill #1a1614
  - b-glow-l [rect] at (110, 0) size 30x1080 fill #3aa7d8
  - b-glow-r [rect] at (1750, 0) size 30x1080 fill #3aa7d8
  - b-body [rect] at (520, 640) size 880x620 radius 240 fill #f1efe9
  - b-collar [rect] at (850, 600) size 220x160 radius 40 fill #2b2622
  - b-head [ellipse] at (800, 200) size 320x400 fill #e2b58e
  - b-hair [rect] at (790, 160) size 340x150 radius 70 fill #151210
  - shot-b-floor [rect] at (0, 880) size 1920x200 fill #0e0b09
- lt-name-b [text] at (140, 954) text "June Halvorsen" 28px weight 700 #ffffff visible [6, 90)
- lt-role-b [text] at (500, 954) text "Member of the Platform Team, Northwind Labs" 28px weight 500 #ffffff visible [10, 90)

## Construction steps
1. Timeline 1920x1080 at 23.976 fps; Fusion composition 90 frames. Local 0 = master frame 1636.
2. Picture track: shot A stand-in (bright grey room, blue window strip, dark-shirted figure) for frames 0-2 with its lower third baked into the same clip so it disappears with the picture; straight cut at frame 3 to shot B stand-in (beige wall, black pillars with a thin blue edge glow, light-shirted figure).
3. Lower third for speaker B: two Text+ nodes, 28 px white on the baseline row y=954, name at x=140 weight 700, role at x=500 weight 500; no plate.
4. Timing: name on at frame 6 (three frames after the cut), role on at frame 10 (four frames later). Both are hard switches; hold to the end.
5. Easy mistake: letting the outgoing graphic linger over the new shot or dissolving it out. It must end on the last frame of shot A exactly, then the new one rebuilds from nothing.

## Content variants
Default copy:
- nameA: Sayid Okonkwo
- roleA: Member of the Platform Team, Northwind Labs
- nameB: June Halvorsen
- roleB: Member of the Platform Team, Northwind Labs
Adaptation changes:
- nameA: Priya Raman
- roleA: Research Lead, Northwind Labs
- nameB: Tomas Eriksen
- roleB: Product Engineer, Northwind Labs

## 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.
