Real databending and channel displacement, not a preset overlay. Push an image until it breaks, then dial it back until it looks right.
A glitch is a picture of a system failing. Channel shifts come from analogue video losing sync, blocky displacement comes from a compressed frame referencing data that never arrived, scanline tearing comes from a display refreshing while the buffer is being written. Deliberate glitch art borrows those specific failures because they are legible: people recognise a broken signal even when they could not name the mechanism.
Four mechanisms account for most of what reads as glitch, and they come from genuinely different sources.
Channel displacement, where red, green and blue are offset from one another, comes from analogue colour video where the three signals travel separately and arrive out of step. It is the most recognisable glitch and the easiest to overuse.
Block displacement, where rectangular regions of the image jump sideways, is a compression artefact. Video codecs describe a frame as a set of blocks copied from an earlier frame with a motion vector; corrupt the vectors and blocks get copied from the wrong place. This is why the displaced regions are always rectangular and always aligned to a grid.
Scanline and row shifting comes from horizontal sync failure, where the display starts drawing a row before it knows where the row begins. The offset is per row and usually varies smoothly, which is why it reads as a wobble rather than as noise.
Quantisation banding and colour bleeding come from bit depth reduction and chroma subsampling, the same processes that let video fit down a wire in the first place.
Random damage does not look like a glitch. It looks like damage. The difference is that real failures are structured, and the structure is what the eye recognises.
Displacement should align to a grid, because compression works in blocks. Channel shifts should be consistent across the whole frame or across coherent regions, because the signal delay is global. Row offsets should be correlated between neighbouring rows, because sync error drifts rather than jumping.
The other half of the illusion is that a glitch has to be partial. An image where every pixel is broken carries no information about what it used to be, and the tension in glitch art comes entirely from the gap between the intact image and the damage. Leave most of the frame readable.
Repetition also matters. Real signal failures recur: the same block corrupts on successive frames, the same scanline tears. Damage that appears exactly once reads as an accident rather than as a system in trouble.
Glitch stacks well with the other deterministic effects, and the ordering changes the result substantially.
Glitch before dithering gives you damaged tone that then gets screened, so the dither pattern runs continuously across the broken regions and the whole image reads as one degraded object.
Dither before glitch gives you a clean screened image that has then been corrupted, so the damage sits on top of an intact structure. This is usually the more legible of the two.
Adding scanlines and a slight bloom afterwards pushes the result toward CRT rather than digital file. Adding chromatic aberration at the frame edges pushes it toward lens and camera. Which one you want depends on what kind of failure you are claiming.
Most glitch tools are built on random number generation, which means every run gives a different result and none of them can be reproduced. That is fine for a one-off and useless for a series.
Bitgrain's glitch effects are seeded and deterministic. The displacement pattern is a function of the settings, so the same image with the same parameters produces identical output every time. You can build a look, save it, and apply it consistently across twelve covers or across every frame of a video without the damage flickering between frames.
That last point matters more than it sounds. Frame-independent random glitch applied to video produces a strobing mess, because the damage changes completely sixty times a second. Deterministic glitch lets you hold a corruption steady for as long as you want and then move it deliberately.
It imitates channel displacement from analogue video, where the red, green and blue signals travel separately and arrive slightly out of step. Offsetting the channels reproduces it.
Because it is unstructured. Real failures are block aligned, row correlated and partial. Keep most of the frame intact, align displacement to a grid, and repeat the damage rather than scattering it.
Yes, and because the effects are deterministic rather than random, the damage stays stable between frames instead of strobing. That is the difference between a glitch that holds and one that flickers.
Yes. Every glitch effect is seeded and deterministic, so a saved project reproduces exactly, which is what you need when the work is a series rather than a single image.