Everything about immersive projection for a 100+ audience concert you wanted to know but never bothered to ask
We just pulled off a small private modern piano concert for a 100+ audience with immersive projection across two walls at 90°, driven by 3 projectors.
The show worked. The audience was pulled in.
Here's the stuff I wish I'd known before starting.
Aimed at people doing small to mid-size immersive concerts and performances without a giant budget, a giant crew, or six months to prototype.
-------------
The space
1. You don't need 360°. You need 140°.
People hear "immersive" and imagine projection on every wall, the ceiling, the floor, the bartender, maybe the dog outside. You probably don't need that.
Human binocular vision is ~120°. If the visuals fill 140° from where the audience sits, the experience reads as fully immersive.
Don't light the ceiling - or the floor - or the wall behind the audience - unless it REALLY matters.
Save the lumens for what people actually stare at.
2. The corner is a curse AND a blessing.
Our space used two walls meeting at 90°.
That geometry hates normal flat panoramic content. If you throw one ultra-wide image across the corner, it can look like a giant TV bent in the middle. From any seat that is not perfectly centered, the seam becomes obvious.
But the same corner becomes powerful if you let it carry meaning.
It can be the corner of a real-feeling room, the edge of an aquarium you're diving into, the divider between two train windows. Treat the corner as a feature, not a seam to hide.
3. Mount projectors with overlap, correct in software.
Two projectors lighting adjacent walls geometrically want to be mounted in the same spot — which is usually impossible.
Don’t try to tile the projectors perfectly in the physical world. It becomes a trap. Even projectors from the same manufacturer will have slightly different brightness, color, lens behavior, and geometry.
Mount them practically. Give yourself some deliberate overlap. Then fix the geometry in mapping software and blend edges to have smooth image across all surfaces - corner included.
For drop ceilings, look into scissor clamps / T-bar clamps. They grip the ceiling grid without damaging the tiles and are fully reversible. For a temporary event, this matters.
4. Cover the walls with cheap white/grey fabric.
Wall paint has color cast, texture, and patches you'll spend hours grading around. For a one-night event, ~$70 of photo-backdrop muslin (white or light grey) and a bag of corkboard pins gives you a uniform surface in 30 minutes.
Left - cheap photo backdrop. Right - a wall painted dark green.
5. Contrast beats resolution.
Audiences read contrast before they read pixels. A high-contrast 720p image at venue scale looks better than a flat 4K one.
If the image is muddy, low-contrast, and washed out, the audience won’t care that it was technically 4K. Remember - immersive videos augment performance and create atmosphere.
The Content
6. Some music doesn't want visuals.
Dense technical music — complex jazz, virtuoso solo work — fills its own perceptual space. Adding visuals competes for attention without adding meaning. Cinematic, slow-decay, atmospheric music is what visuals help. If you're choosing repertoire, choose music with room in it.
We got lucky here: moody, cinematic, atmospheric piano music by Alex Syedin that genuinely benefits from projection.
7. Visuals need a thread of story.
Yes, visuals are mesmerizing. Mesmerizing. Humans are helpless in front of moving light.
But there is a difference between visual stimulus and visual experience.
Strobes, abstract geometry, fractals are visual magnets that the audience won't remember as anything more than "there were flickering lights."
Even a thin narrative — a season change, a journey out of Earth — makes the visuals stick.
For each piece, define in one sentence: “What is happening here emotionally?”
If you can answer that, the visuals get easier and better.
8. Audiences either watch or they listen. Not both.
This is the most counterintuitive lesson. Anything cinematic — camera movement, faces, tight sync to musical beats — pulls people into video-mode and out of music-mode. For immersive accompaniment to live music, you want conditions that drift, not events that announce themselves. Camera movement is the loudest "watch me" signal cinema has.
For live music, avoid visuals that constantly say “look at me.” Create a visual state the audience can live inside while still listening.
9. Three projectors in a row means you’re effectively making a 16:3 show
This is where many people first discover pain.
Three horizontal HD projectors behave like an ultra-wide canvas with aspect ratio 16:3 (3 × 16:9 = 48:9 = 16:3). Sounds simple until you try to make content for it:
- many tools cap at 21:9
- no AI tools were trained for it
- composition tends to collapse into the center
- normal cinematic framing does not work
- normal stock footage looks wrong or empty
It's tempting to author three independent 16:9 videos and route one to each projector. Don't. In practice, that is brutally hard to author, synchronize, revise, and perform live.
One ultra-wide master timeline is dramatically easier to author, sync, and revise.
10. Let the audio drive timing.
Two important aspects here.
First, live concert visuals are not the same as a music video.
In a music video, every beat can flash or change. In a live concert, that can become exhausting fast. If the visuals react to every accent, the projection starts competing with the music - not what we want for an immersive video that supports the music.
Second, trusting your ears to find peaks and valleys may be suboptimal.
Intuitive approach is to listen to a piece, get a feel for its shape, and time visuals to where the music feels like it builds, peaks, and rests. This is wrong more often than you'd think — what you hear and what's actually in the audio aren't the same thing. Your ears smooth things over.
Drop your audio file into Audacity (free) or any DAW. You're looking at two things:
- Waveform view — shows you peaks and quiet sections at a glance.
- Spectrogram view — shows you where high-frequency content lives vs. low. Useful for catching things like "this section has crystalline high notes, this one is all low rumbling" — which often maps to whether visuals should feel sparse-and-bright or dense-and-heavy.
Here is the link to the actual piece - for reference.
The quiet tails at the ends of pieces - which is also high-pitched in this example - may be the biggest visual canvases of the show, and most planning ignores them.
Try planning from the data - then compare the impressions.
-----------------------------------------------------------------
The Tools
Projectors
For small spaces, you want something simple, compact, with 2000–4000 lumen output. Short throw is essentially mandatory — otherwise musicians get lit and cast shadows on your projection.
UST (ultra-short-throw) may be cost-prohibitive. Used options on eBay may be noisy, bulky, or come with lamps near end-of-life. Skip anything labeled "home theater" or anything overly smart — you don't need notifications about a Google account mid-performance, firmware prompts, auto-keystone surprises, or sudden re-calibration because the projector decided so.
Simple. Reliable. Enough light. Right resolution. Basic geometry correction. Quiet. That's the whole list.
We ended up renting 3 x Optoma GT2100HDR - perfect fit for our needs.
Connecting projectors to your computer
Trivial question, right? HDMI cable, I suppose?...
Remember, this project LOOKS easy.
Connecting one or two projectors is trivial.
On the third you may find you've run out of ports and need a USB hub. If on Mac, you'll then suddenly discover that DisplayPort MST extended mode isn't supported on macOS — USB hubs can only mirror your desktop, not extend it.
In addition, you may find that:
- your laptop just cannot drive that many extended displays
- your GPU output limits are lower than you thought
- your cable run is too long
- one projector negotiates a different resolution
- the operating system rearranges displays for fun
Many projectors now support Ethernet/WiFi streaming and there are exotic options like attaching a Raspberry Pi to each one — these will all keep you busy for a while.
If you want it to just work: long HDMI cables, hidden behind ceiling tiles. Done.
But remember the length! Rule of thumb for 1080p over HDMI is about 15 meters (50 feet), and even shorter for 4k signal. Beyond that length, you're in the world of NDMI extenders and optic cables.
One more gotcha: there's no standard way in Windows or macOS to play a single video across several monitors unless those monitors are virtually joined into one surface by the GPU (e.g. NVidia Surround). For Surround to work, monitors must be connected directly to the card — USB hubs are not supported.
The good news: for projection mapping you don't need Surround. Projection software handles the multi-output split.
Pick your projection software and accept the pain.
Here is a decent shortlist for multi-projector setups.
- MadMapper
- HeavyM
- Resolume
- Isadora
- Millumin
- VDMX
- Disguise (if you are in a different budget universe)
- TouchDesigner.
HeavyM is easier to start with. Resolume is performance-friendly. TouchDesigner is insanely flexible if you are willing to build.
But here is the catch: none is documented at a level any reasonable engineer would call acceptable. Expect to lose hours to counter-intuitive things that should just be written in the manual (yes, MadMapper, we're looking at you**)**.
Pick early, commit, comb through tutorials, learn it deep, stop comparison-shopping.
Most of these tools have restricted trial versions that are good enough for learning. For the actual show, get a temporary license if available. MadMapper, for example, has rental options.
Also: do not reduce yourself to torrents. Have some self-respect. Rent the license, borrow a license, find a student with an educational one if budget is the issue.
A MIDI controller is not optional.
Driving a multi-projector rig with a trackpad in front of an audience is a uniquely stressful experience.
Especially if the performer talks between pieces, changes timing, pauses, skips, extends, or simply breathes differently than the recording.
You need physical buttons.
Use a MIDI controller.
Not later. From the start. Build your show around live cue control even if you think everything will run in a fixed sequence. Live performance always drifts.
A $50–150 MIDI controller (Akai APC Mini, Korg nanoKONTROL) maps directly into projection software and gives you tactile control that's night-and-day better than clicking around a laptop in a dark booth.
At minimum, you want buttons / pads / keys to:
- play
- stop
- pause
- trigger next scene
- trigger previous scene
- slow down / speed up
- reverse / back up (to loop a section if a musician gets carried away with a solo)
- jump to specific chapter cues
- Holding loops
- Fade to black
Pre-map your scene cues to physical buttons. Label them with gaffer tape if needed. Your future self in the booth will thank you.
Creating the 16:3 video
First, pick a resolution and commit. 1080p (1920×1080 per projector → 5760×1080 total) looks great but may choke laptops driving three outputs. 720p (1280×720 per projector → 3840×720 total) is often indistinguishable at venue scale, much easier to author, and stock footage is dramatically easier to source at that height.
Chopping 4K stock footage down to a 5760×1080 strip leaves you with low effective resolution per crop anyway. Don't overshoot.
Surprisingly, both CapCut and DaVinci Resolve support ultra-wide timelines — though you may need paid features (stabilization, optical flow) to make AI-generated source usable.
Remember: with projection, content clarity, contrast, and emotional fit may matter more than theoretical pixel count.
Stock footage
- Pexels.com and Pixabay.com — actually free, decent quality, 4K available.
- iStock, Shutterstock, Storyblocks, Artgrid, Pond5 — paid; royalty-free in the licensing sense, not the "free of charge" sense.
- For YouTube downloads: YT-DLP is a de-facto standard.
Remember: normal stock footage is usually 16:9. When you crop it into 16:3, you may lose most of the useful frame. A beautiful 4K shot can become a narrow strip of sky, floor, or empty nothing.
AI generation of ultra-wide 16:3 video
This is real pain. Most AI video tools weren't built for that shape, top out around 21:9, and don't understand your show as a whole.
The workflow that worked for us:
ChatGPT — surprisingly good for thinking through scenes, structuring a visual arc, generating prompts for other tools, and producing 16:3 stills. It won't make videos, but it'll hold context and generate keyframe ideas that stay coherent across a chapter. Stills won't be 720px tall — that's what upscalers are for.
Midjourney — a schizophrenic painter. Doesn't understand story, loves paint. Ask it for a "bonfire" and it may give you something absolutely deranged. With references, strong prompts, and clear start/end logic, it can do miracles. Accepts --ar 16:3 and produces 16:3 video.
https://reddit.com/link/1thct42/video/b9gt94cjy02h1/player
The most useful workflow was:
- Define the emotional arc of the piece.
- Generate or design 16:3 keyframes, upscale if necessary
- Use those key frames as start/end frames in Midjourney videos - you get coherent 5–20 second 16:3 chunks.
- Slow them in post to fill the timeline (this is where you need paid functions like optical flow and stabilization).
- Stabilize if needed.
- Crossfade, layer, mask, and loop.
- Assemble into one ultra-wide master timeline
Those 15–20 second chunks are valuable. If the motion is slow and atmospheric, you can stretch them, blend them, and make them cover much more time.
But remember one very important limitation of all AI generative tools: they are artists, not cameras.
They are designed to create and change. They are not designed to hold a shot perfectly still while modifying only one tiny part of the scene. They have no camera control. Holding the frame still while only a fraction of the scene evolves is a challenge with no reliable, reproducible recipe. Plenty of tutorials suggest prompts that "may" work; none give predictable results.
Which means your AI-generated clips will probably have some combination of:
- jitter
- unwanted camera movement
- background inconsistency
- object morphing
- weird lighting shifts
- details that change when they should not
- seams that become visible on large projection
Plan for cleanup time. Stabilization in post is your friend.
Optical flow has a hard floor.
Resolve, Premiere, and CapCut all do optical-flow interpolation for slowing footage. At 50% they look identical. At 25% the differences emerge. Below ~15% they all collapse — there's no source data for the algorithm to invent between frames. Plan your source clip lengths accordingly: a 5-second clip will stretch to 10 seconds cleanly, to 20 seconds painfully, to 30 seconds not at all.
Test on the actual hardware as early as possible.
A render that looks gorgeous on a laptop will reveal corner-seam artifacts, banding, and horizon-line creasing the moment it hits three projectors meeting at 90°. The day-of discovery is the most common failure mode. Get even rough timing onto the real walls early.
The Execution
Rehearse.
Sync between visuals and music has to be eyeballed live before an audience sees it. The performer needs to know what's in their peripheral vision, when transitions happen, where visual climaxes land relative to their playing. One full run-through catches 80% of what would have gone wrong. Two catches almost everything.
------------------------------------------------
That's the working set. Every item on this list became a problem because someone (often me) assumed it would just work. None of them just works. Plan for each one.
