Viewer · Good Tools

One model, every way of drawing it. Import a model (GLB/GLTF/STL/OBJ/PLY/3DM Rhino) and pick anywhere in the taxonomy of graphical projections: MULTIVIEW (plan, elevation, and true first- and third-angle sheets — four panes at one scale, mirrored cells), AXONOMETRIC (isometric, dimetric 1:1:½, trimetric, or free orbit), OBLIQUE (cavalier, cabinet, military — with the receding angle, depth ratio, direction and plan roll all live), or PERSPECTIVE (1-, 2-, 3-point and a real curvilinear fisheye, rendered through a cube map so straight edges genuinely bend). Every type carries a description and names the algorithm behind it, and a live read-out measures the view you are actually looking at — foreshortening as you orbit, vanishing points as you tilt. Draw it in a display mode off the Inker's shelf. Exports a PNG and a Rhino Python script that rebuilds the same projection in your own model. Client-side, nothing uploaded.

Model

GLB, GLTF, STL, OBJ, PLY or 3DM (Rhino). Drop a file on the viewer or click Import — nothing is uploaded, it's parsed in your browser. A .3dm needs render meshes saved in it (shade the Rhino view before saving); a .gltf with external textures won't resolve them — prefer .glb.

Projection

Parallel › Oblique › Cabinet

Depth drawn at HALF length, receding at 45°. Halving the depth has no optical justification at all — it is a straight fudge — and it works: the drawing stops looking stretched and starts looking like an object. The name comes from the furniture trade, which drew cabinets this way because it showed the front true and the sides believably.

Convention: ISO 5456-3 — oblique (cabinet) projection

True-shape face

The face kept flat-on and true shape. Put an oblique on the plan and you have a planometric — the military projection's own family.

Oblique

Receding angle

Depth ratio

1.00 is cavalier (depth at true length), 0.50 is cabinet. Nothing about the eye picks either: it is a drawing convention, and you can set any other.

Receding direction

Display

Display mode

The Inker's display modes, on your own model — paper, ink, silhouette and matte fill. YOURS is whatever you last built there. A mode set to layer colors falls back to white here: an imported mesh has no layers.

Edge outlines

Ground grid

Viewport — cabinet

Demo massing

True shape on the front

Depth drawn at 0.50 of true length, receding up and right at 45°.

The flat-on face is true shape and true size; only depth is a convention.

Loading 3D…

Demo massing loaded — cabinet oblique. Pick a projection, or import a model.drop a model · pick a projection

Every projection, and how this tool draws it

A graphical projection is a decision about the sight lines. Keep them PARALLEL and the drawing can be measured but cannot be seen; let them CONVERGE on an eye and the drawing can be seen but not measured. Everything below is one branch of that first choice.

Parallel › Orthographic › Multiview (primary)

The building drawn straight-on, one face at a time: the sight lines are parallel to each other AND perpendicular to the page, so every view is true shape at true size. No single view shows depth — you read depth by comparing views. Everything on the page can be measured, which is why sets of construction drawings are multiview and pictorials are not.

ElevationISO 128-30 · ASME Y14.3 — orthographic views: Looking square at one vertical face. That face comes out true shape and true size — you can measure a window off it with a scale rule — and depth is invisible: everything behind the face collapses onto it. Four of them (front, rear, left, right) describe the outside of a building completely, and none of them tells you how deep it is.; How this tool draws it: Orthographic camera placed on the face's own axis (front = +Z, right = +X, and so on) with world +Y up the page. No shear, no convergence, no roll.
PlanISO 128-30 · ASME Y14.3 — orthographic views: Looking straight down. The horizontal dimensions are true and height is invisible — a two-storey wall and a parapet read the same. This is a plan PROJECTION of the whole massing seen from above; a floor plan is a different drawing, a horizontal section cut about 4 ft above the floor with everything below it drawn.; How this tool draws it: Orthographic camera on +Y, rolled so world −Z is up the page — i.e. the back of the building at the top of the sheet, the orientation both multiview conventions share.
Third-angleASME Y14.3 (third-angle, symbol: truncated cone, small end toward the viewer): Three primary views on one sheet, arranged as if the building sat inside a glass box and each pane were folded outward toward you. The top view goes ABOVE the front view; the right-side view goes to the RIGHT. The rule of thumb: each view lands on the side you looked from. Standard in the United States and Canada, and the free corner (upper right) is where the pictorial goes.; How this tool draws it: Four orthographic cameras rendered into four scissored panes of one canvas, all sharing ONE frustum height so the panes are at the same scale and can be measured against each other. Placement is the only thing the convention changes.
First-angleISO 128-30 · ISO 5456-2 (first-angle, symbol: truncated cone, large end toward the viewer): The same three views — literally the same drawings — but the building is imagined between you and the plane, so each view is pushed THROUGH to the far side. The top view goes BELOW the front; the right-side view goes to the LEFT. Standard across Europe and Asia. Because the drawings are identical and only the arrangement differs, a set that doesn't say which convention it uses can be read exactly backwards; that is what the truncated-cone symbol in the title block is for.; How this tool draws it: The same four cameras and the same shared frustum as third-angle — only the cell each pane is placed in changes (see layoutPanes). Compare the two and the convention is the whole difference.

Parallel › Orthographic › Axonometric (auxiliary)

Still orthographic — parallel sight lines, no convergence — but the object is turned so one view shows all three dimensions at once. Parallel edges stay parallel and each axis keeps a constant scale; what changes is how much each axis is shortened. WHICH axes shorten equally is what gives the projection its name.

IsometricISO 5456-3 — axonometric representations: All three axes equally inclined to the page, so all three shorten by exactly the same amount — 1 : 1 : 1, at 30° and 30° from the horizontal. The true foreshortening is 0.816, but almost nobody applies it: draw all three axes at full size and the drawing simply comes out 22.5% larger, still perfectly consistent. ONE scale for everything is why this is the workhorse — and why it is the one axonometric you can build with a 30/60 set square.; How this tool draws it: Orthographic camera along (1, 1, 1), which makes each axis's cosine to the sight line ±1/√3 and so foreshortens each by √(1 − 1/3) = 0.816.
DimetricISO 5456-3 — axonometric representations (dimetric 1 : 1 : ½): TWO axes shortened alike, the third different. The standard setting is 1 : 1 : 0.5 — width and height at full scale, depth at half — which puts the axes at 7°11′ and 41°25′. Use it when one face carries most of the information and you want it nearly true while still showing depth: the 7° axis is almost an elevation, and the 41° one clearly recedes.; How this tool draws it: Camera direction solved from the ratio itself, not read off a table: fx = fy and fz = r·fx under a unit direction gives a² = r²/(2 + r²), so r = ½ lands exactly on (1/3, 1/3, √7/3).
TrimetricISO 5456-3 — axonometric representations: All three axes shortened differently — 0.94 : 1 : 0.81 here, with the axes at 20° and 30°. Three scales means three rulers, which is why it was rare on the drawing board and ordinary in a computer. What you buy with the trouble is freedom: you can aim at whatever combination of faces the drawing needs rather than accepting the two views a set square gives you.; How this tool draws it: Camera direction inverted from the two axis angles (20° and 30°) via tanθx = sinε·tanφ, tanθz = sinε·cotφ — the same relations the isometric and dimetric fall out of.
Free (orbit)ISO 5456-3 — axonometric representations (general case): A general axonometric: drag to any parallel view you like. It is still orthographic — nothing converges, and each axis still has one constant scale — so this is the family isometric, dimetric and trimetric are the three named cases of. The live foreshortening of each axis is printed below as you orbit; drag until all three read the same and you have found isometric by hand.; How this tool draws it: Free-orbit orthographic camera. The read-out recomputes √(1 − (e·v)²) per axis from the camera's actual world position, about five times a second — it is measuring the view you are looking at, not repeating a preset.

Parallel › Oblique

Parallel sight lines that meet the page at an ANGLE instead of square to it. One face is kept flat-on and true shape, and the third dimension is simply drawn receding at a chosen angle and a chosen fraction of its true length. Fast to draw by hand and untrue to any real viewpoint — no camera anywhere can produce one.

CavalierISO 5456-3 — oblique (cavalier) projection: Depth drawn at FULL length, receding at 45°. Every edge on the page — front face and depth alike — is at the same scale, so the whole drawing can be measured with one rule. It also looks distinctly too deep, because the eye expects distance to shorten things and here nothing does. The name is from military engineering: the cavalier was the raised platform you drew the fortification from.; How this tool draws it: Orthographic camera locked square to the chosen face, with the projection matrix post-sheared so one unit of eye-space depth slides 1.00 unit across the page at 45°. The face is untouched by the shear, so it stays true shape.
CabinetISO 5456-3 — oblique (cabinet) projection: Depth drawn at HALF length, receding at 45°. Halving the depth has no optical justification at all — it is a straight fudge — and it works: the drawing stops looking stretched and starts looking like an object. The name comes from the furniture trade, which drew cabinets this way because it showed the front true and the sides believably.; How this tool draws it: Same shear as cavalier with the depth coefficient set to 0.50 — the only difference between the two projections is that one number.
MilitaryISO 5456-3 — oblique (planometric / military) projection: A PLAN oblique: the plan is kept true shape — turned 45°, so both wall directions read — and the heights are raised straight up the page at full length. It comes from fortification drawing, where the plan carried the real information and the walls only needed to be tall enough to see. Also called planometric. Turn the plan roll to 30° and you have the variant most site axonometrics use.; How this tool draws it: Orthographic camera looking straight down, rolled 45° in plan, with the shear set to 90° at full depth ratio — so eye-space depth (which in a plan view IS height) rises vertically at true length.

Perspective (central)

Sight lines converge on a single eye point instead of staying parallel, so things farther off get smaller and parallel edges run to vanishing points. It is what an eye and a lens do — and it is why nothing in a perspective can be scaled off the page. How many vanishing points you get is decided by how the page sits relative to the building, not by a different kind of projection.

1-pointCentral projection — Alberti's construction, Della Pittura (1435): The page is parallel to one principal face and the camera is level. That face's horizontals stay horizontal and its verticals stay vertical; only the depth edges converge, all to a single vanishing point on the horizon. It is the corridor, the nave, the room interior — and it is fragile: turn the camera even slightly off the face and a second vanishing point appears. Orbit and watch the read-out below change.; How this tool draws it: Perspective camera on +Z looking at the model centre, level and square to the face. The vanishing-point count under the viewport is recomputed from the camera's live direction — a family of parallel lines gets a vanishing point exactly when it is not parallel to the page.
2-pointCentral projection — two-point construction: The camera is LEVEL but turned to face a corner. Because it is level, verticals stay vertical on the page; the two horizontal edge families each run off to their own vanishing point, both sitting on the horizon line at eye height. This is the ordinary building exterior view, and the one architectural renderings default to, because upright verticals read as stable.; How this tool draws it: Perspective camera at the model's own height (no tilt) turned in plan toward a corner. Level is what keeps it at two points: tilt the camera and the verticals acquire a third vanishing point.
3-pointCentral projection — three-point construction: Turned to a corner AND tilted. Now all three edge families converge: the two horizontal ones to the horizon, and the verticals to a third vanishing point below the frame (bird's-eye, as set up here) or above it (worm's-eye, if you tilt up). It is the skyscraper look — and it is what your phone does every time you tip it back to fit a building in, which is why photographs of buildings so often look like they are falling over.; How this tool draws it: Perspective camera both turned in plan and raised above the model, looking down. The third vanishing point is a consequence of the tilt, not a separate construction.
CurvilinearEquidistant azimuthal (fisheye) projection · Barre & Flocon, La perspective curviligne (1968): Straight lines drawn as curves. A flat-page perspective can only hold a limited field before the edges stretch grotesquely — at 120° the corners are already unusable — because the picture plane runs away from the eye faster than the scene does. Project onto a SPHERE instead and the field can reach a full 180° with the distortion spread evenly around it. Every straight edge that does not pass through the centre of the image bends; the amount is set by the field-of-view control.; How this tool draws it: A true fisheye, not a filter: the scene is rendered into a cube map from the camera's position, then each pixel is resampled along the direction the equidistant azimuthal mapping r = f·θ assigns it. Because the rays are traced per pixel, edges curve because the projection curves them — nothing is warping a flat render after the fact.

The ratios and axis angles quoted above are computed from each projection's own camera direction (foreshortening of an axis = √(1 − (e·v)²)), not copied from a table, and projections-selftest re-measures every one of them by projecting unit axis vectors through a real three.js camera. Nothing here is uploaded: the model is parsed and drawn in your browser.