Download the pack. Put it on your phone. Here’s what happened to these files before they got there.
Most glitch art can’t survive contact with a phone.
Not because it looks bad — because of how it’s made. A pixel sort, a channel shift, a datamosh: these are all defined in pixels. Not in inches, not in percentages. When you make a piece at 1080 × 1920 and then scale it up to fit a modern display, every artifact you carefully placed scales up with it. The sharp tears go soft. The scanlines turn to mush. The thing that made it feel like a machine error starts to look like a JPEG someone emailed too many times.
Part of that pipeline is generative. Getting a single composition to hold up across a phone, a desktop, and a watch face means three wildly different aspect ratios — and you can’t just crop your way between them without losing the negative space the piece depends on. We use Luma Labs to generate the correct aspect ratios up front, so each format starts from a composition that was actually built for its frame rather than one that survived a crop. The destructive passes then run per format, at that format’s full resolution.
Which is the same principle as everything above, one level up: the frame is a parameter too.
That’s what the F.26 wallpaper pack is. 3 pieces, exported clean for phone, watch, and desktop. They’re free. No email, no account, no “unlock.” Links are at the bottom.
But since you’re here, here’s what actually happens to a file.
1. The source has to be worth destroying
Glitch is subtractive. It takes information away, smears it, misreads it. Which means if you start with a weak image, you end with a weaker one — noise on top of nothing.
The pieces in this pack start as clean geometry: hard neon outlines on true black. High contrast, generous negative space, a small palette. That matters for a specific technical reason — most destructive effects key off luminance. A pixel sort finds the bright regions and drags them. If your source is flat and mid-toned, the algorithm has nothing to grab and you get an even wash. If your source has a hard edge between a neon line and pure black, the algorithm has a cliff to fall off.
The rule we work by: high contrast in, violent glitch out. Flat in, muddy glitch out.
2. Sorting, and why direction is the whole decision
Pixel sorting is the backbone of most of these. The idea is simple: pick a threshold, find every run of pixels that falls inside it, and reorder that run by brightness. The pixels outside the threshold are the walls — they’re where each sorted run stops.
What people underrate is that the axis is the aesthetic. Sorting vertically reads as gravity — melt, drip, rain, decay. Sorting horizontally reads as transmission — scan, tear, signal loss, a broadcast failing. Same algorithm, same image, two completely different emotional registers. Most of the pieces here are vertical, because a lock screen is a vertical object and the eye wants the smear to fall.
Threshold does the rest — though fair warning, the direction of that control flips between tools. In Asendorf’s original and the implementations built on it, a low threshold means fewer pixels qualify as walls, so the runs get long and luxurious and swallow the composition. Raise it and you get short, surgical streaks that read as deliberate. We live mostly at the high end. The point is to keep the geometry legible underneath the damage: if you can’t tell what broke, nothing broke.
3. Channel separation is where the color multiplies
The cyan-and-magenta fringing on these isn’t a filter applied at the end. It’s the red, green, and blue channels pulled apart from each other by a few pixels and recombined.
Where an offset copy lands on its own, you see that channel raw — a hard red or blue edge. Where two of them overlap, you get a secondary: red and green make yellow, green and blue make cyan, red and blue make magenta. So one line becomes a small spectrum, and the image reads as though it were printed slightly out of register.
Worth being precise about what this can and can’t do, because it’s the thing that determines your source palette. Offsetting channels only redistributes color that’s already in the file. It can’t invent any. A pure cyan line has no red in it — shift it however you like, no red fringe will ever appear. So we build with hues that carry real energy in at least two channels, because that’s what gives the separation something to pull apart. Done right, two colors going in come out as six.
4. The resolution problem, which is the actual hard part
Here’s the thing nobody warns you about.
Say you tune a beautiful sort at 1080 pixels wide. Your smears run about 200 pixels — roughly 18% of the frame. It looks great. Now you want that same piece at 6000 pixels wide for a print or a large display. If you upscale, you get soft garbage. So you re-run the effect on the full-resolution file instead — and your 200-pixel smear is now 3% of the frame. It’s a rounding error. The piece is gone.
Every destructive parameter has to scale with the canvas. Smear length, channel offset distance, block size, scanline thickness — all of it has to be expressed as a fraction of the frame, not a fixed pixel count, or the composition falls apart the moment you change output size.
That’s the entire reason A Glitch House exists as a project rather than a folder of exports. Getting a glitch to hold its character from a 1440-pixel lock screen up to a 24 × 36 print is not a filter. It’s a pipeline.
5. Why a lock screen is the right home for this
One practical note that’s more than aesthetics: these are built on true black, and on an OLED screen a true black pixel is switched off. Nothing is emitted. So the neon doesn’t sit on a dark background — it floats in an absence. The contrast you’re seeing is the display’s actual physical contrast, not a simulation of it.
There’s a small power dividend too, though it’s honest to say it depends on how you use your phone: at high brightness a mostly-black image draws meaningfully less than a bright one, and the gap narrows the further down you keep your screen.
How to set it
iPhone — Save to Photos. Settings → Wallpaper → Add New Wallpaper → Photos, and pick it. Pinch to position, then tap Add. iOS may offer Depth Effect, which lifts part of the image in front of the clock — on abstract work like this it often won’t engage at all, since it needs a recognizable subject to isolate. If it does and you’d rather it didn’t, the toggle lives in the ⋯ menu at the bottom right.
Android — Save to your gallery, then long-press an empty spot on the home screen → Wallpaper and style → choose from gallery. On Samsung, that same screen has a Color palette option that recolors your system UI to match the wallpaper. Turn it off if you want your interface to stay neutral — or leave it on, it picks up the neon better than you’d expect.
Mac — System Settings → Wallpaper → Add Photo, under Your Photos. The fill mode is a separate pop-up menu next to the wallpaper thumbnail rather than part of that flow — set it to Fill Screen. One catch: a 16:9 file on a 16:10 MacBook display gets its sides cropped under Fill Screen. If you’d rather keep the whole composition, use Fit to Screen.
Windows — Right-click the image → Set as desktop background. Or Settings → Personalization → Background, then set “Choose a fit for your desktop image” to Fill.
One tip: these were composed with a lot of dead black at the top and bottom, which is exactly where your clock and widgets live. Don’t fight it — the negative space is the frame. If a piece feels crowded, it’s usually positioned too high.
What’s next
F.26 drops September 5, 2026. Same pipeline, run all the way up — print resolution, properly color-managed, on paper that can actually hold this much black.
If you put one of these on your device, we’d like to see it. Tag @aura_glitch or use #AGlitchHouse and we’ll repost.
Break something today.
