START HERE: the glitch-art starter kit + why 'add a glitch filter' gets you removed
Welcome. Quick orientation so you don't post a Photoshop 'RGB split filter' and get roasted.
Real glitch = you broke real data. Databending (edit the file bytes), pixel sorting (reorder pixels by a threshold), datamoshing (delete I-frames so motion smears), feedback loops (camera into its own output), signal corruption. A one-click 'glitch filter' that simulates the LOOK without touching the data is... fine, flair it honestly, but it's the decaf version.
Free toolkit:
Audacity (open raw image data as U-Law audio, apply effects, export back — classic databend)
a hex editor (HxD / ImHex) for surgical byte corruption
ffmpeg + Avidemux for datamoshing
a pixel-sort script (dozens on GitHub, or roll your own, it's ~40 lines)
AI angle welcome: feeding glitched frames back into img2img, training on corruption, pixel-sorting a diffusion output — all on topic. Just don't let the model do a clean simulation of glitch and call it a databend. We can tell.