PHOTOGRAPHIC MOCKUPS AND SIX VIEWS

The editor's 'Present your object' panel appears after a location renders and
automatically prepares its images when scrolled into view. It does not require
an STL/3MF download or manufacturing solid preparation. Every selected source
tile is loaded and checked before a separate render worker compiles a complete
area snapshot. Camera position and offscreen tile eviction cannot omit buildings
from the mockups. The snapshot uses the shared measured terrain, water planes
and bridge sources, full city geometry and a measured terrain error at most
0.10 mm / one pixel at a 1600-pixel overview. It does not invent map details.
Model size, relief, visible layers and finish changes update presentations.
Frames and full-strength decorative materials remain independent of temporary
navigation quality. Geographic or geometry changes cancel stale preparation.

Presentation preparation has finite budgets: 256 source tiles, 15 million
triangles and 768 MiB of mesh arrays. An area exceeding the complete-snapshot
budget reports an error instead of silently dropping detail. The main streamed
preview and full-detail export retain their separate source and work budgets.

Photographic mode uses four prepared settings: a clean front studio shot,
an angled studio tabletop shot, an oak sideboard and a handheld frame.
Every setting offers white, black, natural pale-oak and dark walnut-brown
frame photographs. Dark wood follows the warm, satin timber reference supplied
for this studio; all four new photographs have individually calibrated openings.
The Frame finish selector changes the photograph while retaining the same
loaded model. No typography appears in individual photographic exports.
These plates were prepared once with the built-in
image_gen tool from photographic references. They are AI-edited backgrounds,
not untouched camera files. No image generation runs when a user changes the
model. Source paths and exact preparation prompts are recorded in
/landing-assets/terrain-photos/provenance.json, studio-provenance.json and
dark-wood-provenance.json.

Each template stores a common top-inner-rim plane, light preset and optional
foreground hand masks in photo-templates.js. The model floor stays at world
zero, below that rim. Angled photographs also store measured outer vertical
frame edges, fitting the height projection independently of the opening.
The complete camera is decomposed into one rigid view and intrinsic projection;
pixel projection, physical depth and material shading share that camera.
A closed invisible frame proxy writes depth before the map and casts shadows.
The near rim hides low model edges, while raised far geometry can cover the
photographic inner wall or rail when it is in front. There is no flat backing
polygon clip or painted aperture shadow across raised geometry. Hand masks
restore the photographed foreground last. The frame remains photographic
where no model geometry is in front of it. Recess and exterior frame dimensions
are estimated from each plate, rather than measured manufacturing dimensions.
Each frame finish has an individually measured photographic rim.
The editor's object Frame setting applies to the six studio renders;
the photographic Frame finish selector chooses the actual photo plate.

The three-dimensional display frame has a bevelled closed surround with border
width and depth both six percent of the terrain base size: 12 mm by 12 mm for
a 200 mm map, giving a 224 mm outer footprint. Its opening preserves the exact
map size. White, black, natural wood, dark wood and no-frame selections update both the live scene and
all six rendered views, including changes made after a mockup has loaded.
Overview cameras include the outer frame; detail views keep their close crop.
Photographic mode hides the three-dimensional surround and uses its calibrated
photo plate. This temporary masking cannot hide the frame in subsequent views.
The display frame remains a presentation asset, excluded from STL/3MF geometry.
Natural oak and warm dark walnut use the same procedural timber materials in
the live scene and six views, with grain following each rail into the mitres.
Grain is measured in model millimetres, filtered at a distance and affects
colour, roughness and lighting only. It does not change the closed surround.
Map print layers, clay finish and navigation quality leave timber unchanged.

Six camera presets provide front, left and right three-quarter, low relief,
centre detail and terrain detail views. Each selected view can be saved as a
1600 px PNG; the six views can also be saved in one 1920 px contact sheet.
Photos use the source photo's aspect ratio. Browser rendering quality is a
deterministic real-time approximation; it does not imply validated printable
geometry or an AI-redrawn model.

Photographic seating leaves source geometry buffers, materials and transforms
unchanged. The depth proxy is owned and disposed by the photographic renderer;
it is hidden in all six model views and never enters manufacturing geometry.
Camera tests reconstruct an independent skewed, off-centre pinhole camera and
verify hardware depth, pixel projection and unprojection. Closed-frame and ray
tests distinguish hidden near backing from visible raised far relief in all
four angled finishes at 160, 200 and 300 mm. All sixteen variants retain native
image, rim, depth, hand protection and invalid-input checks. Real Tower Bridge
browser and PNG evidence is retained under output/research/photo-perspective.