Arendelle

Arendelle is built from three surfaces at three different depths, and the whole design follows from that. A proscenium surround at the plaster line frames the picture and carries the emotional temperature of the show. Four leg panels in the wings on line sets 6 and 12, staggered at 5.9′ and 11.8′ upstage, give the stage depth and a forest to walk through. A cyclorama 29.6′ upstage on line set 30 holds the sky, the ice palace, and every wide vista in the piece.

Three depths means three focus planes, three sets of geometry and three different answers to the shadow problem. Treat them as one big screen and the show is soft on two surfaces out of three. Treat them as three instruments and a modest rig looks like a much larger one.

Every figure below comes off the Corner Canyon house plan: a 50′×21.5′ proscenium opening, 29.6′ of usable stage depth with the cyc on line set 30, a 75′ tech-booth throw, and a TITAN lens 10′ above the house floor — 6.5′ above a deck that stands 3′6″ proud of it. The legs snap to existing line sets rather than inventing positions.

The legs live where curtain legs live, staggered in the traditional masking order: LEG-*1 on line set 6, downstage and furthest offstage, LEG-*2 on line set 12, upstage of it and stepped onstage. Their offstage and onstage edges are set so the pair reads as one receding wall with a deliberate slot of cyc between them.

Alignment must be worked in apparent position, not physical. A leg at 11.8′ upstage is foreshortened to 86% of its true offset from centre; one at 5.9′ to 93%. Lay these out with a tape measure on the deck and they will be nearly two feet out to the eye.

Four cues composited onto all six surfaces using the real staging geometry and the predicted foot-lamberts. The artwork is placeholder — generated to carry the right palette, density and compositional weight, not to stand in for a designer. Judge these for legibility, contrast between surfaces and where the eye goes.

Four properties decide whether a scenic surface takes projection well. They are worth stating plainly because three of them are invisible until the projectors come on, and by then the scenery is built.

Screen gain is how much of the light a surface returns toward the audience. Unprimed muslin sits around 0.7–0.8; a purpose-made front-projection surface such as Rosco's Front White is 0.90 with a 170° viewing angle. That is roughly a 12% brightness gain over good muslin for no extra lumens, and the wide viewing angle matters in a proscenium house where a third of the audience is well off centre.

The trade to know about: high-gain surfaces also return ambient and spill light more efficiently. That is fine here because the design already keeps front light off the projection surfaces — see Section 15.

A 62′×29.5′ piece with a 50′×21.5′ hole in it is an awkward object to build, hang, store and transport, and you never need the middle of it. Against a 50′×21.5′ opening the frame band is a header 62′×8′ and two side bands 6′×21.5′. Build those three pieces, assemble them in place, and the hole takes care of itself.

First choice — SEG fabric frames. An extruded aluminium frame with a silicone-edged IFR projection fabric that zips into a channel and tensions drum-tight. They are light, they are dead flat, the fabric is replaceable if it is damaged mid-run, and they are the closest thing to a purpose-built projection surface you can hang as scenery. They cost more than lumber and they are the right call for a show whose whole design depends on register holding for two weeks.

Second choice — hard-covered Hollywood flats. 1×3 on-edge framing with a 1/4″ ply or lauan skin. Stiffer than a traditional Broadway flat and much less prone to racking at this size. The catch is seams: six flats across the header gives you five vertical joints, and under projection a telegraphed joint reads as a line across the frame. Skim the joints with joint compound, sand flat, then prime. Budget real shop time for that step — it is what separates a good-looking portal from an obviously-built one.

Avoid a soft painted drop with a cut-out. It is the cheapest option and it will move. Every time the HVAC cycles or a door opens backstage, your warp is wrong.

Four panels 5′ wide and 25′ tall. The proportions are the problem: at 5:1 these are slender enough to flex, sail and swing, and each one carries an independently warped projection region that a two-inch movement ruins.

Restraint is the whole game. Fly each panel from a batten or spot line with a pipe pocket or hanging irons at the top, and positively tie the bottom — floor plates and stage screws, or a bottom pipe tied off to the deck. Do not rely on weight alone. Actors brush past legs; a panel that can move six inches will, on opening night.

Give the offstage edge of each panel a 2–3″ black border in paint or gaffer. Canvas edges are never straight, and the mask feather in the media assumes a soft edge to land on rather than a hard bright sliver on the wing wall.

A 50′×30′ front-projection surface, and the one place to spend money on material rather than construction. Only 45.1′×25.4′ of it is ever lit (Section 07), but the full cloth still has to be there: the unlit margin is what keeps the frame of the picture off the wing walls and the grid.

• Vinyl or poly cyc, IFR, seamless. Rosco's vinyl cycs run seamless to 40′ with ultrasonically welded joins beyond that — invisible at normal viewing distance and, unlike sewn muslin, they do not wrinkle. Rose Brand's 128″ Poly Cyc IFR is the sewn-panel equivalent and a reasonable budget step down.
• Front White (gain 0.90, 170°) if the budget reaches it. The extra gain and the wide viewing angle both pay off here, and the design never rear-projects.
• Tension it properly. Pipe or chain pocket at the bottom, tied off at both bottom corners, and pulled out at the sides. A hung cyc bellies, and a bellying cyc is why the plot specifies a Bézier warp with an 8×8 control grid rather than simple corner pinning.
• Do not paint a cyc you intend to project on unless it arrives unprimed and the manufacturer says to. Factory finish beats shop finish on this surface every time.

• Rosco Projection Screen Paint on the portal and legs, or a premium flat (not matte, not eggshell) interior latex in a neutral white. Two coats over a sanded, primed surface.
• Roll, do not spray, unless your shop sprays well. Roller stipple is invisible under projection; spray orange-peel is not.
• Test before you commit 1,800 ft². Paint two 4′×8′ sample boards, hang them in the actual positions, and look at them under the actual projectors with actual content. Half an hour here saves a repaint.
• Fire retardancy: use FR paint or FR-treat after painting, and keep the certificate with the show paperwork.

These come off the Corner Canyon house plan rather than from assumption, and they are what the generator in the asset pack is loaded with. Three of them overturned earlier working figures, and two of those changed the plot.

Two Digital Projection TITAN 1080p Quad 2000 and one Epson EB-PQ2216W. Fifty-six thousand lumens in the room, and the rig is lens-limited, not lumen-limited — every lens wants a long throw, which decides the whole plot.

The Epson belongs on the cyc for four reasons that all point the same way: its 0.35:1 lens is the only glass in the kit that can frame a 50′ image from over the stage rather than from the house; it puts 85 px/ft on the surface against a TITAN's 38, on the one surface the audience studies; its laser will not drift in colour across a two-week run; and its native contrast is far better than the TITANs' 2,000:1, which matters enormously on a cyc that spends half the show in near-blackout.

No TITAN can do that job. At a 75′ FOH the cyc is 105′ away, and zoomed all the way in the narrowest image a TITAN makes at that distance is 56′ wide — it cannot frame a 50′ cyc. Worse, at a 6.5′ lens height the shadow reach from out front is effectively the whole stage: an actor anywhere in the beam would print on the cyc at close to full size.

Both TITANs stack on one raster. There is no spare in this kit, which makes the lamp discipline below a genuine operational requirement rather than good practice.

Both are consequences of fixed geometry, both are knowable now, and both have a scenic answer that costs nothing. Neither is a reason to change the plot.

The surround is solved by glass, not by scenery. Your side bands are only 6′ wide to begin with — a 50′ opening in a 62′ canvas — so the stock lens leaving 4′ of each band dark would have gutted the frame rather than trimmed it. The DP 109-236 HB zoom (1.16–1.49:1) run at 1.21 covers the full 62′ from the tech booth. It is the one purchase this plot depends on. Order it early enough to hang with it, not after.

The cyc margin is unavoidable. The unlit strip is 2.45′ at each side, 3.81′ above the image and 0.81′ below, and it cannot be fixed without moving a unit whose position the lens has already decided. Sideways this barely matters: the upstage legs mask the cyc almost exactly to its apparent edges, so the side margin is hidden. The top is the one to deal with. The header covers the cyc down to about 21.5′ apparent, and the image tops out around 20.2′ apparent, which leaves roughly 1.3′ of bare cloth showing in a band right under the header. Hang a black border or a teaser to take it. Do not try to solve it in content — black pixels and no pixels look identical, but bare unlit cloth catches stage-light spill and glows grey.

If you can free line set 13, moving the Epson one set downstage grows the lit area to 47.9′×27.0′ and closes most of the margin. It costs you clearance: 0.99′ between the projector and the leg-2 pipe instead of 1.97′. Worth measuring on the day; not worth assuming.

Both surfaces are close to 2:1 — 50×30 and 62×29.5 — and every projector available is 1.78:1. The surface is wider than the projector, so sizing is width-driven: set the raster width to the coverage width and mask the vertical overspill.

Run that through a side-by-side pair and it falls apart. Two units each covering 33′ of a 62′ surround throw only 18.6′ of height on a 29.5′ surface. Force them to cover the full height and each raster becomes 52′ wide, overlapping by 42′ — an accidental stack with extra steps, which QLab's blend then dims back down. Stack instead: same footprint, double the lumens, no seam, and a dead unit leaves you at half brightness rather than half a picture.

One FOH lens holds the surround at 0′ and the legs at 5.9′ and 11.8′. Blur on the surface is governed by one relationship, and throw distance cancels out of it entirely:

A 62′-wide image off a 0.67″ DMD is about 1,310× magnification, which makes the blur vanishing: 0.07 px at the near legs, 0.14 px at the far ones. Focus on the portal frame and the legs come along free.

The rule this exposes: blur scales as the inverse square of coverage width. Wide coverage means a short lens and enormous depth of field; narrow coverage with a long lens is where focus actually bites. Cover 62′ and a 12-foot depth spread costs 0.14 px; cover 20′ with a long lens from the same position and it costs 1.6 px. That is the quiet reason dedicated long-lens leg projectors are a worse idea than they look.

The 0.35:1 lens sits 15.79′ from the cyc and 24.92′ above the deck — 12.2′ above the centre of its own image. Light hits the top almost square-on from close range and the bottom obliquely from half again as far. Illuminance follows cos(incidence) ÷ distance²:

The gradient is the same 6.5:1 either way — moving the unit one line set downstage buys lit area, not evenness, and costs about 2 fL everywhere because the same light spreads over 13% more surface. Hang at minimum shift, around 24.92′ trim: racking the shift up to mount higher makes the gradient worse, not better, and buys almost nothing in shadow reach. Then decide deliberately how much of the gradient to keep — the asset pack ships two correction masks computed from this exact geometry, one halving it and one flattening it. Flattening pulls the whole cyc to its dimmest reading, so half is usually the right trade. Measure with a meter at five heights at tech before choosing.

Leg-on-leg shadow. The beam heading for the upstage leg has to get past the downstage one, and whether it does is decided at the downstage leg's own plane. An earlier version of this staging had them exactly tangent — no shadow at all, which sounds ideal and is the worst place to be. Two inches of drift either opens a bare slot or drops a hard shadow line down the upstage panel. The current positions carry 2.15 ft of real margin.

Panel behind the proscenium can never be lit. A leg reaching further off centre than the FOH unit can see through the opening has a strip no front projector can reach. From house centre the proscenium hides it and nobody notices. From the opposite side of the house it is a dark stripe down the panel edge — visible to a whole block of the audience and invisible from the tech table. The FOH reach is 26.97′ at 5.9′ upstage and 28.95′ at 11.8′; the legs sit at 26.00 and 20.39. Both checks run on every rebuild of the asset pack and print clearance in feet.

QLab 5's signal path runs Source → Cue → Stage → Region → Route → Device. Build it in reverse — devices first, content last — and nothing has to be redone.

The decision that makes this show work: one stage per surface, not one stage per projector. Your designer thinks in surfaces; the cyc is a 2:1 canvas and a leg is a 1:5 canvas, and those are what a video cue should target. The projectors are delivery, handled entirely at the region and route layer. Change a projector, re-warp one region, and every cue in the show still works.

Greyscale images at stage size: black hides, white shows, mid-grey dims proportionally. Since QLab 5.6 a stage mask reloads live when the file on disk changes — keep Photoshop open at the tech table and iterate the edge without touching the workspace.

Encoding

• HAP or HAP Q in a .mov, not H.264. HAP decodes on the GPU and loops without hitching; H.264 at these pixel counts drops frames the moment three surfaces play at once.
• Every background must loop seamlessly. Almost every cue is an ambient hold of unknown length — the ice palace stays up as long as the scene runs.
• Deliver a matching still frame for every moving cue. If playback fails mid-show a still is a recoverable night; black is not.

Design rules for this rig

• Nothing critical in the bottom 8′ of the cyc. That is the 3.0 fL zone and the shadow-and-bounce band. The previz in Section 03 shows exactly what happens when you ignore this.
• Legs are texture, not picture. At 5′ wide a leg cannot carry a scene. Bark, ice crystal, aurora light, frost creep. Vertical motion only — horizontal motion on a 1:5 canvas reads as an error.
• Grade the upstage legs about a quarter stop hotter than the downstage pair, or accept that they sit back — 12.2 fL against 10.6. They are 6′ further from the lens. Make it a decision in the edit rather than a surprise at tech.
• Colour-match legs to cyc, not to the portal. The audience sees legs and cyc in the same glance through the opening; the portal is a separate visual frame.
• Grade on the actual surfaces at show brightness. Content that looks rich on a laptop goes flat at 8 fL. You will push contrast and saturation well past what looks right on a monitor.

A starting allocation in running order. The recurring principle: the portal is the emotional temperature of the show and changes rarely — roughly nine times across two acts — while the cyc changes with every location and the legs change with the world state.

Two viable paths, and the licensing question is not optional — Disney titles are actively policed and your production licence does not cover projected imagery you source yourself.

Licensed package. Broadway Media sells authorised Scenic Projections for both Frozen and Frozen JR through MTI, pre-cued to the script. It comes in two tiers and the tier matters enormously here: the standard licence delivers playback through their own StagePlayer app, while the professional licence is the one that gives you QLab files and permits mapping and custom elements. For a six-surface mapped rig the standard tier is not usable — buy professional or do not buy it.

Original content. Anything you originate still has to clear with MTI and Disney, and the authors' Frozen licence has design restrictions attached. Ask your MTI rep in writing before your designer starts work, not after.

Whichever path you take, expect to originate the leg content yourself. No stock package ships 1:5 portrait loops cut for four independent panels, and it is the least labour-intensive content in the show to make.

The sequence matters more than the hours. Hanging before the survey, focusing before the scenery is trimmed, or warping before the canvas is tensioned each cost a full day. Six phases, each with a gate that has to close before the next one starts.

Run this once, properly, on load-in day 2. Everything projects from the asset pack — grids, placement guide and grey fields are in assets/templates-pack/grids/.

Everything below is generated from one set of venue numbers by generate.py. Edit the VENUE dictionary with your surveyed measurements and re-run — every grid, mask, template and clearance check rebuilds against the real room, and the script prints the leg warp corners to type into QLab.