Edge blending: getting an invisible overlap between projectors

Edge blending is the technique that merges the images of two or more overlapping projectors into one continuous picture. Each projector fades its brightness progressively across the shared band, called the overlap zone or blend zone. Done right, the seam disappears and the audience never suspects there is more than one machine. Done wrong, you get a bright stripe down the middle of your content.

I have blended projectors for 15 years, from the Arc de Triomphe (15 Barco units on the 2020 edition) to the Museum of Art and Light in Kansas, 108 projectors over 3,400 m². The method below is the one I still follow on every multi projector setup. It has not changed much. The mistakes people make have not changed either.

How edge blending works

Two projectors display images that overlap on a band. Inside that band, the left projector ramps its brightness down toward the right, and the right projector ramps down toward the left. The two ramps are calculated so their sum stays constant. The result: uniform brightness across the whole surface, no visible transition.

Without blending, the overlap band receives light from both machines at once. It is roughly twice as bright as the rest of the image. You cannot miss it, and neither can your client.

The pixel math is worth doing before you order anything. Total canvas width equals the sum of the projector widths minus the overlaps:

• 2 projectors at 1920 px with a 15% overlap (288 px) give 3840 - 288 = 3552 px of usable canvas
• 3 projectors in the same configuration give 5760 - 576 = 5184 px
• every extra 5% of overlap costs you about 96 px per seam at 1080p

The multi-projector calculator does this for any configuration, returns the real canvas resolution, and exports the matching test patterns per projector.

Overlap zone: 10 to 20% of each projector's width

The first parameter to lock. My field ranges:

• 10%: minimum viable. Tight blend, almost no margin for alignment error. Only for stable projectors on a well-controlled installation
• 15%: the live event standard, and the minimum Resolume recommends in its own documentation. Good balance between transition quality and resolution loss
• 20% and above: comfortable and forgiving. Go there if the surface is irregular or the rigging is not rock solid

Below 10%, the transition is too short to hide anything. Above 25 or 30%, you are burning nearly a third of each projector on the seam for a marginal visual gain. Some hardware vendors recommend 25 to 30% by default; on a flat screen with decent projectors, I have rarely needed it.

Blend curve: avoid pure linear

The shape of the brightness ramp matters as much as its width.

• Linear: constant slope. Simple, and almost always visible as a "hard" spot in the middle of the zone
• Gamma (power): the ramp is corrected for the projector's non-linear light output. Two linear ramps that sum to 100% in signal do not sum to constant light on the wall, which is exactly why this correction exists. Standard on most media servers
• S-curve: slow at the edges, fast in the center. The most natural result to my eye. Pick it when your software offers it

In practice the difference between gamma and S-curve is subtle. The difference between either of them and pure linear is not. Resolume Arena, for instance, exposes gamma per color channel plus a power parameter for the curve steepness; most professional tools have equivalents.

Black level: the gray band problem

A projector displaying black still projects light. A very dark gray, but light. In the overlap zone those grays add up, so on dark content the blend band glows slightly brighter than the rest. Your blend can be perfect on a bright showreel and fall apart on the first fade to black.

The fix is black level compensation (sometimes called black level lift): the media server raises the black of the non-overlapped areas to match the doubled black of the blend zone. You lose a little contrast everywhere, you gain a uniform black everywhere. On a projection show, that trade is always worth it.

Second uniformity issue: brightness matching. Two projectors of the same model, same batch, same settings, differ by 5 to 10% in light output. Common even on new units. Adjust each projector's output manually until a full white field looks even, before touching any blend curve.

Same logic for color: two identical machines drift apart in colorimetry, and the blend zone is where the mismatch shows first. On the Culturespaces venues, 60 to 150 projectors per site have to stay coherent in color; a single drifting unit is visible immediately in its overlaps. Match colors on a common reference before blending. This is part of a wider projector calibration workflow, not a blend setting.

Alignment first: warp before you blend

Blending only works on images that are already geometrically aligned in the overlap zone. The order is non-negotiable: physical alignment, warping, then blending. Blend two misaligned images and the curve will happily average them into a blurry, doubled band that no amount of tuning will fix.

It is the number one mistake on the setups I get called to rescue. I made it too, years ago, exactly once.

The surface changes your tolerance. On historic architecture (stone, brick, relief), texture masks small alignment errors; the material absorbs a pixel of drift. On a smooth cyclorama or a white wall, every pixel of misalignment shows. The cleaner the surface, the more perfect the warp has to be. I cover this and eleven other classics in the calibration mistakes that ruin a mapping project.

Where to blend: media server, projector, or processor

In the media server. The default answer on 90% of projects. Full control over curves, black level and color, adjustable in real time during the show, no extra hardware. MadMapper and Resolume both handle it well for small and mid-size setups. On my own large installations the blend lives in Modulo Player or Modulo Kinetic; across 250+ Modulo servers deployed, I have never needed an external blending box. Cost: GPU headroom, especially with many outputs.

In the projector firmware. Most professional projectors ship with built-in blending. Independent from the server and latency-free, but the curves are limited, black level compensation is often crude, and you configure everything through on-screen menus. Fine for a permanent 2 or 3 projector row. Painful beyond that.

In a dedicated processor. External hardware (Analog Way and similar) for large or broadcast-grade systems. Reliable, expensive, and one more box in the signal chain. I reserve it for projects that genuinely need it, which is not many.

My 7-step blending method

1. Size the overlaps on paper. Number of projectors, overlap percentage, real canvas resolution. Before renting anything. The multi-projector calculator covers this step
2. Install and align physically. Position, focus, zoom, lens shift. Then let the projectors warm up 20 to 30 minutes: alignment drifts while the optics reach temperature, and a warp done cold is a warp done twice
3. Warp each projector. Start with a coarse 2x2 grid, refine progressively. Goal: pixel-accurate geometry in the overlap zone, verified with calibration test patterns, grids and straight lines. What must line up, lines up
4. Match black levels and brightness. Full black field, check uniformity, enable black level compensation, then trim each projector's output to match its neighbors
5. Activate the blend. Set the blend width to your physical overlap, pick a gamma or S-curve, fine-tune on flat colors and gradients
6. Verify with real content. White, black, saturated colors, gradients, moving video. Different viewing angles. One full fade to black to validate the black level. If possible, with the HVAC running and the house lights in show state
7. Plan recalibration. On permanent installations, day-night temperature cycles, metal structure expansion and air currents move projectors by a few pixels. Enough to degrade a blend. I schedule a check every 2 to 3 months; camera-based autocalibration helps, a trained eye still decides

A good blend is one nobody notices. The audience never compliments it. That is the point.

When edge blending is not the answer

Some honest cases where I advise against it:

• One projector can do the job. If a single brighter unit covers your surface at the lux level you need, take it. One 30,000 lumen machine beats two blended 15,000 lumen machines: fewer failure points, no seam to maintain, simpler signal chain
• High ambient light. Blending degrades gracefully in the dark and ungracefully in daylight. If the room cannot be controlled, an LED wall is often the better tool, and no blend curve will change that
• No black level compensation available. If your content is mostly dark scenes and your processing cannot lift blacks, the gray band will be visible every night. Fix the tool chain first or reframe the project
• A "quick blend" on a moving rig. Projectors on an unstable truss lose their alignment faster than you can retune it. Solve the rigging before the blending

For the full calibration chain around blending, warping, color and maintenance, start from the projector calibration guide.

To go deeper on the vendor side: Resolume's edge blending documentation is a solid reference for software blending, and Dexon's edge blending overview covers the hardware processor angle.

If you have two projectors, one wall and a doubt, the calculator is free. And if you want to check that doubt with someone who has blended 108 projectors in one building, write me.