The projection planner is a working optical instrument: real lens data, real brightness arithmetic, a venue you can orbit. This manual covers all of it, from the first drag to the printed plan, including multi-screen scenes, the shadow study, the keyboard, the physics that surprises people, and how a tilted machine warps its picture square again.
The cards run in planning order down the page: the picture and its distance (01), the shape in plan (02), the machine (03), its lens (04), solutions on a curve (05), the array (06), and what the light is worth on the surface (07). Controls that depend on a parent setting, the meter panel, the keystone tools, the space limits fields, stay out of sight until that setting is on, then slide in beside it.
1
2
3
4
5
6
7
The dark viewport is a measured 3D venue. Dragging it orbits the camera and nothing else: objects never move by accident, only when a Move mode is on.
The cube in the corner is the camera's compass, in the 3ds Max idiom. It turns with the venue; clicking a face, an edge dot, or a corner dot glides the camera to that view. TOP gives the floor plan, FRONT the elevation, corners the three-quarter views. The Perspective / Isometric switch changes the lens: perspective is a natural camera, isometric is a drawing-office camera where parallel lines stay parallel, so distances compare honestly anywhere in the frame.
Wide view grows the venue over a dimmed page; Pop out lifts it into a window you can drag and resize by its corner grip, the canvas refitting itself to the window at the same field of view. Perspective stays perspective at every size: the camera never crowds in just because the canvas grew.
Turn on a Move mode (The surface or The projector) and three grammars become available at once. Use whichever suits your hand.
With the projector in hand, the venue also draws the available range: a pale blue volume marking everywhere the projector can stand with the lens still serving the surface square, no tilt, no keystone. It is built per machine and per lens from the real figures: sideways and in height from the lens-shift envelope (with the cut corners honored, since full vertical and full horizontal shift are never available together, which is why its cross-section is an octagon, not a box), and in depth from the zoom span at the current image size. It stays up while you drag, faded so it never hides the work, and the Guide lines chip turns it off with the rest of the linework.

Dragging the object moves it in the floor plane: toward and away walks the throw, sideways walks it off axis. Shift + drag moves vertically: the rig rises and falls, the lens riding with it. With the surface selected and the beam decoupled, Shift + drag walks the surface in depth instead.
In card 01, give the surface its own size and it becomes a canvas: a physical screen larger or smaller than the picture, with the image lighting only the section it covers. Picture and canvas then move independently: Move: the surface walks the canvas itself (drag, arrows, or typed coordinates), while the position sliders, the projector, and lens shift walk the picture across it. Overhang is called out in the readouts, and a decoupled beam roams the canvas torch-like, exactly like content playing across one section of a wide scrim. If the picture and canvas drift apart, Resolve to the canvas puts them back together in one click: one centered projector sized to fit, or the right count of blended landscape or portrait tiles filling it edge to edge, rigged on the ceiling.
A move gizmo stands on the active object: X runs across the venue, Y runs up, Z runs along the throw. Click an arrowhead and drag: the move locks to that one axis no matter how the camera is turned. The surface shows X and Y, plus Z when the beam is decoupled; the projector and free-placed units show all three.

Press R or click Rotate the projector and the arrows become rings: the red ring pitches, the green ring yaws, and the blue ring rolls the chassis about its own aim (90° of roll is a portrait mount). Grab a ring and drag along it; the body turns at about a fifth of a degree per pixel, whichever way the camera faces. Right-drag or Alt + drag anywhere on the body does the same without the rings.

The screen itself rotates the same way. With Move: the surface on, the chip reads Rotate the surface (or press R) and the same three rings land on the screen: rake it back, turn it, bank it, ±80° of yaw and pitch, ±180° of roll, or type the angles in the coordinate row. The first turn quietly decouples the beam, freezing it exactly as aimed, because a rotated screen and a married throw cannot both be true: from there the picture lands as the true ray-cut of the beam on the raked face, keystone measured against the screen’s real angle, coverage clipped in the screen’s own plane, and the brightness gradient following the face’s real normal. Hand pins ride the rotated face, and the warp desk shows it square-on, as a commissioning tool would. Reset position squares everything again.

Whenever Move is on, a coordinate row sits under the venue: X, Y and Z in your working units, plus yaw, pitch and roll in degrees where rotation applies. Type a value and press Enter; the object goes exactly there, inside the same limits the drags respect. The row relabels itself for whatever is selected, and it is the precision path: drag to explore, type to align.
Normally the planner keeps the projector and the surface married: move one and the optics follow. Decouple the beam breaks the marriage. The beam becomes a torch out of the body: a real light cone that goes wherever the chassis points, spreading and dimming with distance. Only where it crosses the surface does a picture land; the rest washes the back wall as raw light, drawn dashed.

Once decoupled you can move the projector and the surface fully independently, in depth too, and rotate the projector freely: ±80° of yaw and pitch, ±180° of roll. The readout under the venue tells you which optical regime the aim is in: inside the lens's shift envelope the offset is carried optically and the picture stays square; past the envelope the chassis is physically tilted, and the readout states the tilt angle and what it costs. A stacked rig decouples too: one converged raster, the whole tower turning with the beam, the summed lumens carried through every readout. Only true multi-tile arrays (blend grids and free-placed rigs) stay married to their tiles, because there each unit owns its own beam.
Why the regimes matter. Lens shift moves the picture without touching its shape. Tilt changes the angle the light arrives at, and the picture keystones. Both are legitimate; the planner just refuses to hide which one your geometry is using.
When a projector cannot reach its surface with lens shift alone, it must tilt, and a tilted beam lands as a wedge: wider where the light travels further, brighter where it lands nearer. That wedge is keystone. Squaring it back up is a raster operation, and the field knows three digital routes. The planner models all three, and it only offers the ones the selected machine actually carries.
And one machine to remember: the Sony VPL-GTZ380 carries no digital geometry at all. Its installation budget is a very large shift envelope, and the planner will tell you exactly that when you tilt one.
All three spend the same coin. The correction pre-shrinks the raster inside the panel, so pixels park black outside the picture and their lumens leave with them; the readouts and the printed plan give the percentage spent. The black floor still lands as the whole wedge. Focus softens as correction grows, a warning Panasonic prints in its own spec file. And no route levels the light: correction squares the shape, never the illumination, so the near edge of a corrected picture stays brighter than the far one by the inverse square of the throw and the angle of arrival. The planner draws that gradient on the surface and states the near-to-far ratio. One more honesty for the fully unlocked rig: decoupling frees the projector from the throw triangle altogether, so the planner also carries each lens's nominal focus span (the makers quote roughly 70 to 1,000 inches of picture); park the machine outside it and the picture lands soft, drawn with its defocus and flagged in the readouts.

Hand pin puts the four corners in your hands, the Quick Corner way: drag them straight on the picture in the venue, per unit on a free-placed rig. Warp desk opens the flat commissioning view every warp tool draws: the surface square-on, the raw footprint, the raster and its pins, with the venue's own content or the alignment grid mapped through the true projective chain and lit by the same range-and-angle physics as the venue. Click a pin and type its offsets, nudge with the arrow keys (5 mm a press, Shift for 50 mm), snap the lot back square, copy one unit's pins to the whole rig, or save the offsets as a pin set and recall them later; sets travel with the plan link. Every move is audited live: a corner cannot leave the raw beam, crossed pins are refused, and the raster spend, the light loss and the near/far gradient track the pins as they go.

Machine only / + external processor sets whose rules apply. Machine only enforces the selected projector's own engine, its published stops, and greys the routes it does not carry. Adding the external processor puts a warp box upstream: every route opens on every machine, no menu stop applies, and the raster becomes the only budget, exactly the media-server case.


Rule of thumb. Shift first, rig second, digital geometry last. Shift is free, moving the rig costs a ladder, and every digital route costs pixels, light and focus forever. When you must correct: menu keystone for a one-axis error, the pin for a compound or rolled one, the mesh for a curve.
Card 06, Open the array, opens multi-projector work; Back to one projector in its header closes it again, columns, rows and stack reset to one. The grid picker sizes the mosaic the way you would size a table: hover the cells to preview the shape, click and the array is set, the caption spelling out units, picture size and tile size. The Columns and Rows buttons still step one axis at a time. Either way, columns and rows tile one wide picture from several machines; the overlap slider sets the blend band, where two projectors ramp against each other and sum to one tile's light. Chassis and screen orientation are independent: portrait tiles hug tall walls, and a mismatch parks unused pixels black, with the readouts saying how many.
Freed units answer to the whole scene: swing one toward a neighboring screen and its light lands there, drawn as a warm patch on that screen’s face labeled with the unit that threw it, the unit’s own wall still blocking whatever it catches first.
Stacking converges machines on the same frame for brightness: the 1 / 2 / 3 buttons cover the common cases, and the count box beside them takes any stack up to eight. Towers climb four chassis high; a fifth unit starts a second tower alongside, with a gap between the columns, the way riggers actually build them. Every chassis in the stack throws from its own lens, so the venue draws one cone apex per body, all converged on the shared landing; the light is one picture at the stack's summed lumens, which is exactly what a warped, converged stack delivers. To aim or tilt a stack as one rigid rig, turn on Free placement and rotate it (rings, right-drag, or the yaw and pitch sliders), or decouple the beam: the whole tower rides the beam frame as one piece.

Free placement releases every unit from its plan seat: select one in the venue or step with Previous/Next, then drag, lift, aim and roll it with the same grammar as the single machine. "Aim at its tile" points it home; "Snap all to plan" rebuilds the mosaic. The Blend map chip appears the moment the rig has two projectors, flat grid or curved wall: every unit's coverage in its own hue, the bands layering brighter where they share, each tile numbered so unit 2 in the venue is unmistakably unit 2 on paper. And every free-placed unit carries its own geometry: the Keystone row, the audit and the hand pins all work per selected unit, and "Copy pins to all units" on the warp desk propagates one unit's offsets to the whole rig, tile-relative. Free placement is also where the focus physics follows each unit alone: drag one off its plan throw, past the lens's focus band, and its picture softens on the wall from its own lens-to-landing distance, with the readout saying so.

The blend tax. Inside a band the picture is only as bright as one tile, but every unit's black floor lands across its whole footprint, so each seam carries a second black floor. Fewer, wider blends are easier to align; more, tighter tiles carry more pixels. The curved-wall solutions rank these trade-offs for you.
The Screens row above the tools puts up to four complete plans in one venue. Each screen is everything this manual describes, surface, projector, lens, array, grade and geometry, with its own seat in the space: an X and Z position and a facing angle. 0° faces the audience, 90° stands the screen along the right-hand side, −90° along the left, so two walls at −90° and 90° face each other across the venue, the classic corridor of light.
The numbered buttons choose which screen the cards drive; the others sleep in the venue as ghosts, and clicking a ghost or its name tag wakes it. + Add a screen copies the current plan and seats the copy alongside, so a matched pair takes three moves: add, type the facing, type the seat. Undo, the plan link, the printed plan and the CSV all carry the whole scene: the paper record adds a screen-by-screen summary, and the spreadsheet a screen column with each seat.

Lux heatmap in the Show row paints illuminance straight onto the surface, worked per point from each lens’s true distance, arrival cosine, zoom and blend ramp: lumens over one tile’s area anchor the scale at the plan throw, then the inverse square reshapes it point by point, with a nominal 10% corner falloff standing in for lens rolloff. Peak, average, minimum and min/peak uniformity read under the venue, hot and cold spots are marked on the face, and a legend carries the scale. It works on flat walls, arrays and drawn curves alike.
Raw field / Flat field are two different instruments. Raw is the bare physics above, honest about throw, stacking and offset. Flat field is the calibrated delivery: a projection lens is engineered to land uniform light on its own focus plane (the ANSI uniformity spec), so on a square-on flat wall it reads even edge to edge, and what remains is geometry alone: keystone stretch on tilted or aimed units, depth variation on drawn curves, and the blend ramps. Raw shows why; flat shows what a dialed-in show measures.
The map is anchored on exposure. On auto, the target works out what a clear picture needs right where you stand: reference white, held seven times above the ambient veil this grade returns, so the mark rises with venue light and falls on deeper grades. Where the delivered light lands under it the map fades, washed the way the audience would see it; where it lands over, it runs hot toward red, and the neutral band reads exactly right. The legend carries the needed figure and the readout states how far the average sits from it. Type your own lux or fL to hold a house number instead, or type 0 for the plain absolute scale.
The meter’s settings live in a small panel that slides in under the Show row while the heatmap is on, grouped and labeled: lens model, the gate, the venue, what it reads, and the target; the × in its corner switches the heatmap and the meter off together. Three more switches make it a working instrument. Blended white / Open gate: blended sums each band to one tile’s light, the calibrated picture; open gate leaves the ramps off, the roughly-double band a meter actually reads during alignment. Dark venue / With ambient adds the held ambient from card 07, the incident meter’s real number with the lights on. Lux / fL on this grade switches between light arriving at the face and light leaving it toward the audience, tied to the planner’s calibrated luminance model.
A ghost screen is never dark: its rig draws faint dashed cones from every lens to its own tile, so a four-screen scene reads as four working projections at a glance, not one live wall among gray ones.
Beams answer to the whole scene. Aim a freed unit past its own wall and its light lands where it truly falls: on a neighboring screen it draws as a warm dashed patch labeled with the unit that threw it. When that neighbor stands between a projector and its own picture, the patch reads lands here first · shadow behind and the warnings name the shadowed picture with the fixes: slide either screen, re-aim the unit, or plan around the cut.
The Rig dimensions chip in the Show row draws the numbers a rigger asks for first, straight on the venue: lens-to-lens spacing center to center, labeled once with “every pair” when the row is uniform; each distinct lens height dropped to the floor; and the vertical pitch between lenses inside a stack. These are lens positions, not chassis dimensions or mounting hardware, so the figures hold whatever frame the units ride in.
Clients ask one question about walk-up surfaces before any other: how close can somebody get before they block the picture? The Shadow study chip, off by default, answers it two ways at once. A 5′ 9″ walker appears in the venue for you to drag anywhere on the floor; every lens whose beam they actually stand in casts their true silhouette onto the surface, sized and softened by the same projection geometry as the picture itself, so a walker near the wall throws a small hard shadow and a walker near the lens a large soft one. Stacked lenses each cast their own partial shadow, which is why stacking is also a shadow strategy: the other units half-fill what one body blocks.
The dashed floor line is the summary: where the beam’s lower edge crosses head height. On a ceiling rig the readout states how close a walker can come to the surface before entering the light; on a floor rig thrown up over the audience the clear zone reverses and sits against the wall; and when the beam crosses walking height the whole way, the planner says so plainly instead of drawing a line that does not exist.
The walker moves two ways. Drag them anywhere on the floor, or click them and a small step pad opens at their feet: four arrows that walk the figure in quarter-meter steps, toward the surface, away, left and right, with the cast and the floor line answering every step. Click anywhere else and the pad puts itself away.

A single-row blend no longer has to ride at image height. With the Mount switch on Floor / truss below, the whole row stands on floor stands, every lens low, and the picture rides up on vertical lens shift: chassis square, no keystone, the beams fanning up from the low lenses. The shift readout prices the ask against the envelope, and when it runs out, Raise the projector rig to fit lifts the stands exactly as far as the lens needs. Drawn and curved walls take the same call: switch the mount and the fitted rig steps down from the ceiling truss to floor stands, every unit still square to its tile in plan.

Pick a branded machine and the readouts add On the scale: the maker’s published body weight, without lens, multiplied across the rig, per body, per stacked position and in total, in kilograms and pounds. It is deliberately narrow: bodies only. Rigging frames, projection cages, enclosures, hoists and cabling are not included, and every point load is confirmed against the maker’s current sheet at quotation. Figures marked “about” are nominal, pending a sheet check. The weight rides along in the printed plan and in one column of the CSV.
Switch the shape to Curved and the wall becomes a chain of straight and arc sections, editable in the shape card. Draw the plan opens a drawing board: sketch the wall in one pass, left to right, or start from an L, U, Arc or S template, and the stroke snaps to true lines and arcs with their measurements.
Then refine it like a Bézier tool: drag a point to stretch, pull a diamond handle to bend a section through your cursor, and add a point three ways: the Add a point chip arms one click, a double-click on the line drops one right there, and a selected section offers "Add a point in the middle". Corners can be filleted or chamfered at the size you type. Accept the shape and the venue builds it: the rig moves to the ceiling, the projector count is fitted, and five ranked ways to light the shape appear, from the lean rig to the ultra dense one.

A curved wall is solver-seated: every projector stands square to its own strip of the face, so the Move switch is replaced by the one hand control that keeps the physics honest, turn the wall. Type up to 90° either way and the whole drawn shape rotates in plan; the rig re-seats itself on the turned face, every unit finds its strip again, and throws, focus travel and arrival angles all recompute. Picking a preset or drawing a new wall squares the turn back to zero.

The mouse wheel zooms about the cursor, the span chips set a working scale, and Fit the view brings everything drawn back on screen when a point wanders off the edge. The drawing, the background trace and the venue frame slide together, so zooming never moves a wall in the venue.
Lit from is remembered per screen: side A, side B, or both faces each travel with their own surface through screen switches, kept plans and plan links. Change the side and press Accept and the rig crosses to the face you asked for, even when the shape itself did not change.
With more than one screen in the scene, the board becomes the venue’s shared plan. The surface in hand draws solid; every other screen waits as a pale dashed line with its rig marked, exactly where it stands in the venue. Draw the new wall among them and Accept seats it right where you drew it, position and angle included; moving a whole selection and accepting re-seats the wall the same way. The Surfaces chips above the board, or a click on any pale line, change hands, and + New surface adds a screen with the pencil already in yours.
The wall does not have to start under your pencil. Upload a plan on the drawing board, or a file dropped straight onto it, takes three kinds of source. An SVG is sampled along its longest path and fitted into sections exactly like a freehand stroke. An ASCII DXF from any CAD tool is read entity by entity, lines, polylines with their bulges, arcs and splines, chained end to end where their tips touch, and the longest chain fitted. A picture, a photo or a scan of a plan line, lands on the board as a backdrop; the planner traces the strongest single dark line it can read and fits that, and when no clean line can be read it says so and leaves the picture under your pencil, so drawing over the backdrop is always the fallback that works.
Two habits make imports land well. First, imported files carry no trustworthy units, so type the real developed length into the face measures and the whole shape, backdrop included, scales to match before you accept. Second, expect to edit: a closed outline arrives whole, and you Remove the sides you do not need; a wobbly photo trace is a starting chain of points and handles, not a verdict. Shoot pictures square-on, dark line on light paper; a binary DXF needs saving as ASCII (R12 or later) first, and a PDF wants exporting as an image before it comes in.
Some drawings are honest dead ends: a hairpin with its legs a few feet apart, lit from inside, puts the opposite face between every lens and its tile, and no lens or warp shoots through a wall. The planner says so instead of pretending: those beams draw red in the venue and the warning opens with "Not solvable as drawn", naming the free depth against the lens’s minimum throw. The ways out are the real ones: light the shape from the other face, open the legs, or plan a separate rig per leg. In multi-screen scenes the same honesty applies between screens: a screen standing inside another rig’s beam catches the light first, the patch is labeled and the warning names the shadowed picture.
Accepting a drawn wall is the start of the work, not the end. The venue builds the shape, moves the rig to the ceiling, fits a projector count, and then hands you three levers and a ranked list. This is how to work them.
In card 05, after the lens, the planner ranks five valid ways to light the shape with the chosen machine, the fitted rig first, alternatives after: from the lean rig with the fewest units to the ultra dense one. The figures carry the argument, so read them the way an integrator would: pixel density against how close the audience stands, per-tile brightness against the ambient you set, focus travel against the sag of the curve, and the seam count against alignment labor on a ladder. Use this applies a row to the venue; nothing is committed until you like what lands. When two rows read close, pick the one with fewer seams: pixels are bought once, blends are aligned every load-in.
The count is the master trade. One unit fewer means wider tiles: fewer seams, more focus depth on the arc, but lower pixel density and each machine's lumens spread across more face. One unit more means narrower tiles that hug a tight curve, higher density and brightness, and one more seam to align. Fit the count for me sets the count so each tile is exactly what one panel paints at the current band height and overlap; start there, then push one step each way and watch the warnings.
Follows the wall traces the rig along the curve, each unit standing off its own tile no further than the local center allows: compact, and the beams stay short. Cross shoot stands each unit at full throw on its tile's normal, so on tight curves the beams cross the venue and buy throw length: brighter edges, better focus on deep arcs, but the cones fly through the space, so check them against scenery, trussing and follow-spot lines before committing. The per-unit table notes when a unit is held at its lens's minimum throw: the corner turns tighter than the machine can stand back from, and the cross-shoot layout is usually the cure.
Curved rigs blend along vertical seams only; every tile runs full height. The overlap slider prices the bands: wider bands are kinder to align on a face that curves through them, but each carries the second black floor, so deep grades matter more as bands multiply. Turn on the Blend map the moment the rig has two units: each unit's coverage in its own hue, the shared bands reading brighter, tiles numbered so unit 3 on the wall is unit 3 on paper. The Panther test image is the honest check: if the seams read milky on it at your ambient, they will read milky on show black.
The lit from chips on the drawing board choose the side: Side A where the audience stands, Side B from behind the wall, or Both, the Transparent Scrim setup where each face carries its own image. Lighting from behind flips concavity: what you drew as a bowl lights as a dome, and the solver re-seats the rig on the far side.
Band height and base-off-floor size the lit band itself; tall bands flip the fitted chassis to portrait automatically, narrow footprints hugging the curve. Stacks work on curves exactly as on flat walls: the count box adds brightness per station without touching the tile geometry. And when the venue's geometry will not meet the wall square, turn the wall (chapter 07) rotates the whole face in plan and lets the solver re-seat everything.
| What you see | Why it is correct |
|---|---|
| Zooming wider makes the picture dimmer | Zoom holds the distance and changes the picture size. Brightness follows area, not distance: the same lumens over four times the area is a quarter of the light. |
| Moving the projector closer does not brighten a held picture | With the image size held, the lens zooms to compensate: same lumens, same area, same light. |
| A tilted unit is brighter on one edge | Inverse square at work: the near edge of a keystoned footprint can land visibly more light than the far one. The per-unit table quotes the ratio. |
| Blend seams look milky on dark content | Two black floors ride in every band. The Panther test image shows it plainly; deeper grades hold it down. |
| A brighter surface grade reads worse in a lit venue | The grade returns the venue's light along with the projector's. The ladder runs your machine across every grade at your ambient level so the trade reads in numbers. |
| A stack of eight is one picture, eight times the light | Stacked machines converge on one raster, so lumens sum while size, pixels and keystone stay those of a single unit. Columns and rows grow the picture instead; only stacking grows the light. |
| A rotated screen keystones a square beam | Keystone follows the angle between beam and surface, whichever one moved. Rake the screen 20° and a shift-squared throw lands exactly as keystoned as a 20° chassis tilt onto a square wall. |
| A unit dragged off its throw goes soft | A lens holds focus over a finite span. Park a free-placed unit past it and the picture lands, but soft: the planner draws the defocus and says so. |
| The front row is further than expected | Two limits: the pixel grid must have dissolved (about 3,400 × the pixel pitch) and the picture should stay under 60° of width. The seating readout applies both. |
Stand in front of it. Every row above is something you can watch happen on a real surface at the Hollywood demo facility. Bring your plan link; request a demo and we will run your plan at your ambient, on your grade.
The two tables below also exist as the key card: one printable, ink-friendly page for the desk beside the planner.
| Input | Does |
|---|---|
| Drag | Orbits the camera. With Move on and the object grabbed: moves it in the floor plane (surface: in the wall plane). |
| Shift + drag | Moves the grabbed object vertically; decoupled surface: walks it in depth. |
| Right-drag or Alt + drag | Rotates the projector body: yaw with x, pitch with y. |
| XYZ arrowheads | Drag one: the move locks to that axis. |
| Rotation rings | Drag along a ring: pitch (red), yaw (green), roll (blue), about 0.2° per pixel. On the projector or on the surface, per the Move switch. |
| R | Toggles the rotation rings on the Move target: the projector (with free placement or a decoupled beam) or the surface itself. |
| Click the walker | With the shadow study on: opens the step pad; its arrows walk the figure in quarter-meter steps. Drag still works; clicking elsewhere closes the pad. |
| Pin handles | With Hand pin on: drag a corner of the picture; a move that leaves the raw beam, or crosses its neighbor, is refused. |
| Warp desk: arrows | Nudge the selected pin 5 mm; Shift takes 50 mm; Esc closes the desk. |
| View cube | Click a face, edge or corner: the camera glides to that view. |
| Coordinate fields + Enter | Sets exact X, Y, Z, yaw, pitch, roll. |
| Ctrl Z / Ctrl Y | Undo and redo the whole plan, forty steps, camera untouched. Cmd on a Mac. |
| Input | Does |
|---|---|
| Draw a stroke | Left to right; on release it snaps to true lines and arcs. Faithful / Smoother / Simplest set how closely the fit follows. |
| Drag a point / diamond | Stretches a section / bends it through your cursor. |
| Shift while dragging a point | Snaps its leg to 45° steps: square corners fall into the hand. |
| Double-click the line | Adds a point right there. The Add a point chip arms one click for the same thing. |
| Shift + click points | Gathers several; a drag then moves them together. |
| Right-drag / Shift + right-drag | Box select / lasso select. |
| Arrow keys | Nudge the selection; Shift takes bigger steps. |
| Delete | Removes the selected points; their neighbors join. |
| Z / Y | Undo / redo the sketch. Esc closes the board without touching the venue. |