impasto

side business · shipping an end to end product · 2026 to now

Impasto sells wallpaper packs for phones and desktops: photographs rebuilt as digital oil paintings with thick palette knife ridges, linen texture, and wet shadows, then cropped for every screen size. Ten paintings per pack, no subscriptions.

from photo to painting

Photograph of a brown bear standing on a rock in a river
the photo (Cal Rollins, Unsplash)
The same bear rebuilt as a digital oil painting
the painting that ships

The composition survives. The medium does not.

how a pack gets chosen

Painting something beautiful that nobody wants is the easy way to lose a weekend. So the pipeline decides what to paint before it paints anything, and that decision is the part I am building machine learning for.

  1. collect signals

    search volume, wallpaper subreddits, and stock-photo demand, scraped into one table of what people actually want on a screen.

  2. rank by predicted demandmachine learning

    a model scores candidate subjects and compositions against those signals, so the catalogue is chosen by evidence rather than by my taste.

  3. repaint the winners

    only the top ranked subjects get the image-model repaint and the quality pass. Nothing is painted on a hunch.

  4. bundle and sell

    ten paintings become a themed pack, cropped for every device, launched to an audience the ranking already proved exists.

recent work

Digital oil painting of A-frame cabins with solar roofs in a snowy forest at duskDigital oil painting of a motorboat on an alpine lake under cumulus clouds

the stack

  • pipeline/ Python: select the subject, repaint it with an image model, run a quality pass, then crop for every device aspect ratio
  • site/ Next.js, TypeScript, and Tailwind storefront with checkout via Lemon Squeezy

The rest is product discipline: honest copy (they are digital paintings, and say so), licensed source imagery only, and pricing simple enough to explain in one line.

status: store launching soon