Pawtraits turns your pet into a star with AI-generated portraits - themed sessions that drop your dog, cat, or bunny into a holiday, a trip somewhere, or a famous painting. It was one of the first apps Josh and I built under Tortuga Labs, our consumer app studio, and we built it for two reasons. One was to get our hands on the newest image generation models and learn what they could do. The other was to build something we could invest in up front and then mostly leave alone: a product that stays valuable to people over time without much maintenance from us. Plus, my dog Theodore hates outfits so with Pawtraits, I get to style him without the barking.
Features
You start by training a model on your pet - eight to ten photos is enough - and from then on you pick whichever pack you want and generate. A pack is a theme: Halloween, Beach Days, Paris, Famous Paintings. Each generation gives you ten portraits of your pet inside that world, matched to its features, and you favorite the ones you love. There are 28 packs across three shelves: holidays, travel, and pop culture.
Your favorites flow into a public feed. We added it for people who don't have pets but still want something nice to scroll, and it doubles as a gallery for anyone deciding whether to try the app - real output, real pets. We started calling it the happiest feed on your phone. We live in pretty crazy times, and there's something genuinely good about opening an app and seeing nothing but cute animals in funny outfits.
Strategic decisions
- Pay per pack, not a subscription. Pawtraits launched on a subscription, and we moved off it when image generation costs dropped. Charging for the packs you actually want fits a kitschy, fun app far better than a recurring fee does. Subscriptions earn their keep when a product is part of your day-to-day; for something you reach for around a holiday, paying for the thing in front of you is the honest model, and it converts better too.
- Built to run itself. This was the real bet, and the most useful thing the project taught me. We made enough packs - all of them pegged to a season or a moment - that we could schedule a full year of notifications in advance. A week before Halloween, users get a nudge asking what their pet is dressing up as. Same around the holidays, the Fourth of July, whatever's next on the calendar. Once that engine was set, Pawtraits mostly didn't need us. We touch it when a meaningfully better image model ships and clears our testing, and otherwise it keeps delivering without updates. The notifications carry the whole thing: we see roughly a 70 percent open rate off them and a sharp jump in pack purchases right after one goes out.
- Less costume, more pet. We capped accessories at a hat or a collar, with full getups limited to a scene or two per pack. Pile on the costume and two things break: the generation gets unpredictable, and the pet stops looking like the pet.
Learnings
The biggest lesson was that "set it and forget it" is a real product strategy. Most of what we build needs constant feeding. Pawtraits proved that if you do the up-front work right - enough content, a calendar that maps to people's lives, a notification engine tied to it - you can build something that stays alive on its own. That changed how I think about which projects are worth which kind of effort.
The smaller, craftier lesson was that the quality rules came from the failures, not from planning. Early packs leaned on full costumes and the results were a mess - the pet vanished under the outfit. We pulled accessories back and standardized how every scene was built, and the quality jumped. None of that was obvious before we'd generated a few hundred bad images.
Pawtraits is fun, and we still open it to scroll the feed. Sometimes a cute little app that runs itself is exactly the right thing to have built.
Download Pawtraits on the App Store.
