Datamosh a photo into broken blocks

Your picture breaks into compression blocks that jump, smear and bleed, then gets remapped to four bold inks, like a video file that copied the wrong frame. No AI.

Open the editor with this effect

What you get

Made with plain math, not a model

There is no generative model behind this look and nothing was trained on anyone's work. It is a short list of fixed steps running in your browser.

  1. Read the photo: Your photo is read pixel by pixel inside your browser. It is never uploaded, so nothing leaves your device.
  2. Broken blocks: The image is cut into compression style blocks that slip, smear and drop out in a seeded pattern, like a video frame decoded with missing data.
  3. Same input, same pixels: Every step is fixed math with fixed seeds. Run it again tomorrow on the same photo and you get exactly the same image.

Templates in this style

Browse templates in this style

Datamoshing started as a happy accident in broken video files. A clip loses a frame it depends on, and suddenly blocks of the old picture slide and smear across the new one, colours bleeding where they should not be. Artists began doing it on purpose. Bitgrain gives a still photo that same broken, blocky look with plain image maths, no AI and no upload.

What real datamoshing is

Video files save space by storing a full picture only now and then. These are keyframes. The frames between them only describe how small square blocks of the picture move.

If a keyframe is deleted, the movement instructions get applied to the wrong picture. Blocks from one shot drag across the next, faces melt into backgrounds and pixels bleed in the direction things were moving.

The same block structure shows up when a file is damaged or a stream drops data. You see squares in the wrong place, smeared rows and patches of flat colour.

Music videos and album art made the look popular in the late 2000s, and it has stayed a shorthand for anything raw, digital and slightly out of control.

What the effect does to a still photo

It cuts the photo into a grid of blocks, like the blocks a video codec works with, and fills many of them from somewhere else in the picture, so pieces jump out of place.

Some blocks smear, lifting towards the brightest ink, and some drop out to dark patches, the way missing data looks on screen. A few are outlined with a dashed edge, like a broken selection.

The colour channels are pulled slightly apart inside the blocks for a touch of RGB split, and a little noise sits over everything.

Finally the whole picture is mapped onto four inks from dark to light. The default runs from deep violet through purple and orange to cream, and you can pick your own or keep the true colours.

Photos that suit it

Portraits and close crops keep enough of the subject recognisable through the damage, which is what makes a glitch read as a glitch.

Night shots, concerts and neon scenes look at home, though very dark photos can turn into mostly shadow ink. Raise the brightness first if that happens.

Try a few Layout numbers until the dark blocks fall somewhere interesting, rather than right across the eyes.

How to do it in Bitgrain

Open the editor from this page. The Datamosh style is already on the sample. Drop in your own photo.

Set Block size for how big the broken squares are, and Scatter for how many of them break. Turn Ink mix down to zero to keep your photo's true colours.

Under Advanced, Jump distance sets how far blocks travel, Smear and RGB split add bleed, Noise adds grain and Strength brightens or darkens the result. Layout picks a new pattern.

Also under Advanced, Shadow ink, Low ink, High ink and Light ink set the four colours, with Contrast, Brightness and Invert below. Export PNG when you are happy.

Common questions

Can I datamosh a video with this?

This page works on still images. It recreates the look of a datamoshed frame, so it suits covers, posters and posts. The video editor has its own glitch effects for clips.

Will I get the same result if I open it again?

Yes. The break is seeded, so the same photo with the same settings always gives the same blocks. Change the Layout number when you want a new pattern.

Does it use AI?

No. The blocks are cut, moved and coloured by a fixed algorithm, in your browser. Nothing is generated and nothing is uploaded.

Related tools