Sixty-two runs through one plain logo
A flat black-and-white mark, rolled sixty-two times. No photograph, no gradient, nothing soft anywhere in the frame - and every run still came back different.
A logo is the least generous thing you can hand an image tool. A photograph arrives with grain, a tonal range, edges that fall off softly - something for a treatment to get hold of. A flat mark has none of it: two colours, hard edges, and a lot of white doing nothing.
So it is a fair question, and someone asked it. What actually happens if you feed one in?
That is the whole input. Nothing staged, masked or prepared beforehand. It went in as a plain file on a white background and got rolled.
Roll it and look
Explorer picks its own effects and lands its own parameters on every roll, so none of what follows was chosen by hand. The bias was left wide open, no family locked off. On a photograph that setting produces mush. On a mark it does the opposite: with so little in the frame, nothing has anywhere to hide, and you can see precisely what each treatment is doing.
Colour has to come from somewhere, and there is none in the source. So anything chromatic here has been generated - channel offsets, compression fringing, the halo an effect leaves behind when it splits a hard edge. On a photograph you read that as damage to the picture. On a mark it reads as the mark itself being printed badly, which is a more useful thing to have.
Some rolls stop treating it as a logo altogether and start treating it as a tile. Same input, scattered across the canvas - closer to wrapping paper than to branding, and useful for exactly that.
Every so often a roll lands on something that looks entirely deliberate. A knocked-out badge on a halftone field is a real design decision, arrived at by dice.
And every so often it lands somewhere the mark barely survives. Those are worth keeping in the sheet rather than deleting. The failures tell you where the edge is, and the good ones sit just inside it.
That one made the trip worthwhile: solid shapes turned to outline, sitting on black. Nothing in the input suggested it was in there.
What this is actually for
Not a brand system. Nobody should ship a logo that changes every time it is rendered.
But a mark is the cleanest test material there is. If you want to know what a treatment really does - where it puts colour, how it handles a hard edge, what it does with empty space - take the photograph away and run it on something flat. Whatever comes back is the effect itself, with nothing to hide behind.
The second use is closer to the point. Sixty-two runs is a couple of minutes of clicking. Somewhere in that sheet are the stickers, the end frames, the loading screens and the animated version of a mark that until then only existed as one flat file.