Point it at a photo and get a crisp pixel art avatar. It finds the face, frames it, lifts the subject off the background and redraws the lot in a handful of colours. All of that happens on your iPhone, and nothing is uploaded anywhere.
iPhone and iPad · iOS 17 or later · Free · No account
PixelThat works end to end and we are testing it on real photos. The App Store is the next stop, and it will be free there, with no ads and nothing to buy inside it.
Quick facts about PixelThat for iOS
What is PixelThat?
Pixel art looks simple, which is exactly why shrinking a photo does not work. Scale a face down to 32 pixels and you get mush. So PixelThat does what a pixel artist would do. It finds the face and frames it the way an avatar is framed, separates the subject from the background so the edges stay clean, flattens the colours to a small deliberate palette taken from the picture itself, and only then lays down the pixels.
What you can do with it
Auto framing puts the head where an avatar's head belongs. Or drag the selection box yourself. Move it, pinch it, tap to jump, and the result follows.
Transparent, a flat colour, a two-colour gradient, a pixel-art blue sky with clouds, or keep the photo. The cutout has a threshold, grow and shrink, and a live preview when it needs a nudge.
A palette built from your own picture, or PICO-8, Sweetie 16, Endesga 32, Game Boy, C64, CGA, ZX Spectrum, greyscale and 1 bit. You can also paste in your own hex colours.
Natural, Soft, Bold, Comic, Retro, Arcade, Game Boy, Neon and Mono set the palette, dithering, grade and ink together. Everything they change stays editable.
Every picture you keep is filed with the settings that made it. Tap it later to put those same settings on a different photo. That is how you get a set of avatars that match.
No account, no analytics, no network. Your photos are read, turned into pixels and put back down again, all on your phone.
Taken on an iPhone, straight from the app
The short version
PixelThat collects nothing. There is no account, no analytics, no advertising identifier and no networking code in the app at all. Your photos are read, turned into pixel art and saved back again, all on your own device.
The short version
PixelThat is free to use, for personal or commercial work, on Apple devices you own or control. The pictures you make are yours. The app is provided as it is, without warranty, and it is up to you to have the right to use the photos you put into it.
Questions, bugs and requests
Everything goes to hello@cs42.org. There is no account to recover and no data to restore, so most answers are quick. Usually a day or two.
No. The app has no networking code: it makes no requests, to us or to anyone else. You can put the phone in airplane mode and everything still works.
No. There is no sign-in, no email to give and nothing to create.
The cutout uses Apple's on-device subject detection, which is very good but not perfect on busy scenes. Open Background not right? Adjust the cutout and lower the threshold to keep more of the subject, raise it to remove more, or grow and shrink the edge. Switching the model to Portrait (people) helps with hair.
32 pixels is the sweet spot for a face. At 16 you keep the shape of the hair and eyes and it reads as a sprite. At 64 and above the expression survives. Use the tight zoom for the smallest sizes.
Make the first one, tap the bookmark button to keep it, then open the next photo and tap that picture under Recents. Every setting comes back, so the set matches.
In the app's own storage on your device, not in your photo library and not in iCloud. More settings › App › Clear kept pictures removes them all, and deleting the app removes everything it has ever stored.
Only what you ask it to. Save to Photos writes the enlarged copy, which is what most sites want as an avatar. Share hands over both the native pixels and the enlargement without touching your library.
Most websites resize a 32 × 32 file and blur it in the process. The enlarged copy is the same pixels scaled up with hard edges, so it survives being scaled again. Upload that one.
iPhone and iPad running iOS 17 or later. Face detection and subject cutout use the device's own vision models.
Not at the moment. There is a desktop version that runs locally on Windows, macOS and Linux, so write to hello@cs42.org if you would like it.