# Open the report, reframe, then disclose navigation

**Draft construction guide. Native Resolve operation and moving-source fidelity
are not verified.** This is one candidate, with two editable content variants.

## Timing and source

Official film identifier: `2gCqVb2lBwk`. Source interval
[3510,3780), 60/1 fps.
Create a silent 1920×1080 composition with 270 included frames,
local 0..269. Source ordinal=3510+local frame.
Local timeline duration=4.500 seconds. Do not
substitute 24/25/29.97 fps or append a closing hold. The hash-bound preserved
research excerpt is 960×540; native timing comes from its historical source ledger,
not a guessed rate or fresh full-master decode.

## Inputs and setup

Unpack the ZIP and run `python3 verify_package.py` before changing anything.
`project/` is a complete TypeScript/Remotion project. `inputs/default.json` and
`inputs/adaptation.json` hold the two invented content sets. The bundled Geist
font includes its license. Install it for Text+ and calibrate pixel sizes and
baselines against the rendered candidate. The chosen palette is original, not
an official brand token. No source movie, narration or source screenshot belongs
in the editable composition.

`node --experimental-strip-types scripts/controls.mjs` regenerates
`docs/controls.csv` from the same integer-frame model as Remotion. It contains
one row per included frame and only this treatment's animated controls. For
Fusion use global frame=composition start+CSV frame. A scalar step is a held key,
not a default spline. For a continuous value, either key every supplied frame or
recreate the declared smoothstep and compare all rows. Apply each local transform
before its parent. Native Resolve import/export and pixel parity remain untested.

## Fusion layers and geometry

| Group | Pixel geometry | Ownership |
| --- | --- | --- |
| Camera root | origin (195,125) | Uniform scale plus screen-pixel translation |
| Window | x=192, y=120, w=1535, h=844, r=16 | Child of camera root |
| Conversation report card | window-local x=432, y=206, w=700, h=432 | Clipped before mode switch |
| Full document | x=365, y=75, w=750, h=1600 | Appears at document mode |
| Citation rail | x=1222, y=48, w=313, h=796 | Appears with document mode |
| Contents popup | x=84, y=78+contentsOffsetY, w=314, h=741 | Later overlay, not reflow |

1. Keep the conversation and embedded report through 65. At 66 hard-switch to
   document mode: full document and right citations rail appear together. The
   outer window and background persist. No paper-rise animation is invented.
2. Keep cameraScale=1 through 102. Smoothstep to 1.59 at 140 around root point
   (195,125); simultaneously cameraX=0→−9 and cameraY=0→−12. Translate in screen
   pixels after scaling: screenPoint=origin+scale×(worldPoint−origin)+translation.
   This clips the citations rail at the right edge; do not delete it to imitate
   the crop. The document itself is not independently enlarged.
3. Contents remains alpha 0 through 163. Smoothstep to 1 at 171 with local Y offset
   12→0. It overlays the existing left gutter/document edge after the camera
   has settled; the document is not newly allocated a pane at this event.
4. Document:750 px width,60 px padding; title 31 px/1.2,650 weight; section headings
   23/25 px,600 weight; paragraphs 19 px/1.45. Each invented paragraph contains an
   original citation marker. Before the mode switch the shared document subtree
   is scaled.82 and clipped inside its card.
5. Contents uses 14 px eyebrow,19 px title and 18 px section labels. Four section rows
   begin at 176+127×index. Pointer values are window-transformed root coordinates:
   subtract Window's x=192, y=120 before placing the cursor inside the window.
6. The cursor uses held authored positions, not a recovered source trajectory.
   The source's later contents browsing is not reproduced. Preserve the held
   document with visible contents to 269; add no section jump or report export.

Mode/camera/navigation anchors are authored within inspected source sample
brackets, not measured product event latencies. Exact typography, cursor travel,
source chrome, popup opening and late contents-list scrolling remain differences.


## Verification and limits

Compare both actual Remotion exports, not just their manifests. Check the first
and final included frames, event brackets and clipping domains. The source movie
is a research comparison only. The original artwork and this schematic UI do
not demonstrate real product behavior. Mechanical passes do not close the
published source gaps. No native .drp, automatic Resolve import or performance
claim is supplied. All timing interpolation between sampled anchors is authored.
