Nine art styles. Under 60 seconds per portrait. No signup. Those are the specs we shipped with Pet Imagination, a free AI pet portrait generator we built at Inithouse. Here is what the numbers actually looked like during the build.
Why we built it
We run a portfolio of small products at Inithouse, each one an experiment looking for traction. Pet portraits sat at an intersection we liked: a clear use case (pet owners want fun art of their animals), short generation time (keeps the feedback loop tight), and zero friction (no account, no payment to try).
The bet was simple. If someone uploads a photo of their dog and gets a Renaissance-style portrait back in under a minute, they will either share it or come back. Both of those signals are easy to measure.
The style problem
We started with three styles. Renaissance, Watercolor, and Anime. Early testing showed that style count matters more than style polish. People clicked through every option before picking one. Three felt thin. Nine felt like a real menu.
The final set covers Renaissance, Watercolor, Anime, Sketch, Sheriff, Wizard, Astronaut, Final Boss, and Blocky. Some of those are deliberately odd (Sheriff? Final Boss?) because the weird ones generated the most shares in our testing. A golden retriever in a sheriff hat gets forwarded. A golden retriever in watercolor gets saved quietly.
Processing time
Our target was 60 seconds from upload to finished portrait. The actual median lands below that. Most generations finish faster because we front-load the model call while the user is still picking a style. By the time they confirm, part of the work is already done.
We could push it lower, but faster is not always better here. People who wait 20 seconds for a result trust it less than people who wait 45. There is a sweet spot where the wait feels like the tool is doing real work. Below that, it feels like a filter.
No signup, and why
Every product in our portfolio starts with no signup. It is not a philosophical stance. It is a measurement strategy. Signup walls hide your real usage numbers behind a conversion rate that has nothing to do with the product itself. We want to know: does the thing work? Does anyone care? Signup answers neither question.
For Pet Imagination, removing the wall meant we could see raw generation counts from day one. No guessing how many people bounced at the email field. No inflated "registered users" metric that means nothing when 80% never come back.
What we actually track
The numbers we watch are generation count (how many portraits get made), style distribution (which styles get picked), and return visits (same browser, different day). We do not track individual users because there is no account. Browser fingerprinting would let us, but the privacy trade-off is not worth the data.
Style distribution told us things we did not expect. Watercolor is the most popular overall, but Astronaut has the highest return rate. People who pick Astronaut come back and try other styles. People who pick Watercolor usually generate one portrait and leave. The novelty styles pull people deeper into the product.
What worked, what did not
What worked: the zero-friction entry. Someone shares a link, the recipient uploads a photo, gets a portrait. No account wall breaking the chain. This is the growth loop we wanted and it functions exactly as designed.
What did not work as well: SEO for "AI pet portrait generator" is competitive. Established players with years of backlinks sit on the first page. Our content strategy is slow. Posts like this one contribute a signal, but it takes months, not weeks.
We also underestimated how many people try the tool with photos that are not pets. Landscapes, selfies, food. The model handles them, but the results are unpredictable. We decided not to restrict uploads. If someone wants a Renaissance portrait of their lunch, that is fine.
Current state
Pet Imagination is live at petimagination.com. Free to use, nine styles, no signup required. It is one product in the Inithouse portfolio. We run several of these experiments in parallel, each one testing a different market and interaction pattern.
The product is stable. The question now is distribution: can we get enough people through the door to see whether pet portraits have legs as a standalone product or whether the tech (fast style transfer, no-signup flow) is more valuable folded into something larger.
We will know more in a few months. For now, the numbers say people like making their pets look like wizards.
Top comments (0)