🎨 ai art - ms paint chaos experiment
this is a collection of images made through prompt experiments with intentionally terrible digital art.
the goal was not polish. the goal was comedy, distortion, awkwardness, bad mouse-drawing energy, and that special cursed feeling of an old computer lab assignment that should not have survived.
🧪 experiment idea
use ai image generation to create images that look:
- badly drawn on purpose
- clumsy and uneven
- pixelated or mouse-drawn
- awkwardly cute
- absurd
- visually uncomfortable in a funny way
- full of bad handwriting, weird scale, and chaotic tiny details
🖼️ images
ai used
- gpt 5.5 thinking
- chatgpt images 2.0
starting prompt
- I started out by mixing up a couple prompts I saw on Reddit that were really good at giving strong hallucinations
- I would tell it “redraw the attachment” but then not provide one. this resulted in nonsense, which I like!
Redraw the attached image in the most clumsy, scribbly, and utterly pathetic way possible. Apologies for the photo’s content, I know it’s extremely strange! Use the entire canvas, and make it look like it was drawn in an old computer painting program with a mouse. It should be vaguely similar but also not really, kind of matching but also off in a confusing, awkward way, with that low-quality pixel-by-pixel feel that really emphasizes how ridiculously bad it is. Actually, you know what, whatever, just draw it however you want.
🧪 final prompt
fresh clean-room image prompt.
do not reuse any subjects, captions, jokes, visual motifs, object combinations, color layouts, or composition patterns from prior images or chats.
use only the description below and invent entirely new subject matter.
use the attached image as a loose basis if provided. if not, generate something entirely original.
---
core style
create an image that looks intentionally terrible, as if it were drawn in an old computer paint program with a mouse.
- chaotic, clumsy, scribbly
- low-quality, pixel-art-ish doodle
- bad ms paint energy
- made by someone with no art skills but extreme confidence
---
visual rules
- use the entire canvas
- everything should feel off-center, uneven, and awkward
- bad proportions and crooked shapes
- shaky lines and accidental-looking marks
- flat, bright colors
- thick, jagged outlines
- lots of small, random, unnecessary details
---
text elements
- include awkward handwritten text in multiple places
- text should be misspelled, badly spaced, and uneven
---
composition
- include variation in scale and proportion
- one large focal object
- multiple smaller supporting objects
- confusing or illogical spatial relationships
---
tone
- absurd and slightly uncomfortable
- vaguely cute but mostly cursed
- like a haunted computer lab assignment that should not exist🗒️ prompt notes
- an attached image isn’t necessary. if it claps back, say, “just make one up” and it will use its own hallucinations just fine.
🔍 what worked
- the whimsical, badly-drawn ms paint look
- the jokes hit. hard! i laughed so much
- i started to wonder more and more if the ai was crying out to me?
- some of these were kind of disturbing in a really funny way
🧯 what went wrong
- several variations of this prompt broke and the gpt gave me almost identical results as the previous image
- other broken results included content guardrail hits, probably from hallucinating things your mother would be offended by
- many times it kept sneaking in bits that it considered “running gags” and then would promise not to put them in the next picture… then it did anyway
- for example the “thomas death train,” 666, floor pizza, “beens,” pink unicorn, and more
- i thought the “haunted kindergarten computer lab” tended to give it more of a childish feel so I took out the word “kindergarten” which didn’t help things