Staging maps, position images, scale sketches

A position image shows one fixed spatial fact a shot depends on: where people stand, where a car stops, a path, a size against a person. All of them are internal, and every staging map and position image is made in GPT Image, labelled or not (decided). Staging maps are [best available, not yet validated].

A position image. Option: use it when a position is critical, or after a spatial failure [single strong source]. Add it to the plan for that shot.

• The drawing: thin lines on a white ground, every figure drawn as an outline in a colour of its own, with no shading.

• The view: the shot's own camera view (a front view) for people; top-down for objects, paths and overhead shots. Sources split between the two for people, so the front view stays the default [best available, not yet validated].

• A scale sketch is a simple pencil height comparison of two figures, or of a figure and an object.

• In video it is attached last, and its role line gives it one job (placement, path or scale) and tells the model not to draw its lines, outlines or colours (see Video with Seedance → Attaching references).

• The colour legend. Option: use it when this film's plan asks for it [single strong source]. Ask for it in the plan where gaze or height matters, or where several key lines must be told apart. Each figure's head becomes a circle in a second colour where gaze or height matters, and each key line (a path, an edge, a ridge, a stop line) gets its own colour. You write the legend in a LEGEND section of the prompt and repeat the same legend in the video role line ("the red outline is the grandmother; the yellow circle is her head; the green line is the parapet"), so the map itself needs no words.

• A labelled map of the place. Option: use it when the staging is complex (several people, a route, a chase) [single strong source]. It joins that place's anchors: a top-down plan with the landmarks as simple shapes, short English labels beside them, the camera positions the plan uses as small triangles, and a scale bar. Make it from the approved master plate and the plan, never from a storyboard. In video its role line says it owns layout, positions and distances only, and not to render its lines, labels or colours. Arabic never appears on a map, and the map never goes on the anchors presentation.

Settings and prompt style: see Still prompts, storyboard frames and diagrams → GPT Image's two jobs and its version rule.

Template: a staging map (written for this guide)

A staging diagram, {front view from the shot's camera / top-down}, {aspect}: a flat line drawing on white.
CONTENT: {each figure or object as an outline in its own colour, with its position by screen side and depth}. {A path as one coloured line, if any.}
{LEGEND, only where the plan asks for the colour legend: each figure: outline colour; its head: circle colour; each key line: colour.}
STYLE: thin black lines for the set's main shapes only; no shading, no textures.
LOCKS: no labels, no numbers, no letters, no arrows; nothing else in the frame.

Example: the rooftop, front view (written for this guide)

A staging diagram, front view from the shot's camera at chest height, 16:9: a flat line drawing on white.
CONTENT: a small square table in the lower left, near the camera, with one short cylinder on it for the lantern. The grandmother as a red outline, standing just right of the table, near the camera, facing frame-left toward the table. Two relatives as blue and green outlines, standing close together at the parapet in the far middle. One relative as a purple outline, seated on an upturned crate at the far right. A low parapet wall across the whole frame behind them. A water tank as a tall rectangle at the far left.
STYLE: thin black lines for the parapet, the tank and the table only; no shading.
LOCKS: no labels, no numbers, no letters, no arrows; nothing else in the frame.

Example: a labelled map of the place (written for this guide)

A labelled map of the rooftop, top-down, 16:9: a clean line drawing on white, for a film crew.
INPUTS: Image 1 is the approved master plate of the rooftop: it owns the layout and the landmarks only; draw none of its light, colour or texture.
CONTENT: the rectangular roof; the stair hut in the north-west corner; the water tank against the west parapet; the table near the middle; the washing line strung east to west across the south half; two camera positions as small triangles, one by the stair hut facing south-east, one at the east parapet facing west.
TEXT: exactly these labels, in plain capitals beside their shapes, and nothing else: "STAIRS", "TANK", "TABLE", "WASHING LINE"; a scale bar marked "2 m".
STYLE: thin black lines; landmarks as flat grey shapes; no shading, no textures.
LOCKS: no other text, numbers or letters; no people; nothing else in the frame.

Sources

• @adilinthewildtempo (blog) and Higgsfield's studio (the front view, coloured outlines, telling the video model not to draw the marks); kaye.creatives, Oakley | Meta (the colour legend); Listen to Meh by @nue_nue (project) (the labelled location map).