What I learned from a 14 month failure in ai + print-on-demand
stack spoiler: replicate, nextjs+chakra+vercel, stripe, posthog, and gelato's print-on-demand API.
Link at the end...
In late '22 when ai was just kicking off, I woke up one night in flurry with one clear idea: "personalised kids books, but with images that actually look like the kid"
That feeling of certainty melted into stubborn persistence, weighed way down with doubts about my ability to execute or die trying.
I had a few sales and successes, spent a wild amount of time doing such dumb shit as:
- translating my no-visitor website
- advertisting but changing my ad plan every couple of days
- changing the website often
- instead of talking to customers
I also did a lot of things that didn't work, but weren't so dumb...
- printing flyers
- talking to people I know
- talking to indiehackers instead of more suitable customers
Fourteen months in, that project - making personalised picture books with ai and print-on-demand - is kinda limping along. It's still wizard of oz, there are still manual steps for every order. It does a tiny bit of business but I don't push it because a) I'm feeling pretty burned out on it and b) the existing processes won't scale. It's toil now.
A week ago I started work on a much simpler project: upload one pic of yourself (or whoever), and get a canvas print of that person in the style of Girl with a Pearl Earring. A week in, it's already completely automated... if someone orders while I sleep (hey, it could happen) the printer will start on it before I wake up.
I resisted the urge to add more features, more product options, all that saucy distracting joy. My devops heart wants an overblown event-driven wet dream, but it's just a hot mess of callbacks for now.
More controversially, perhaps, there's no database at all. I dunno if I'll regret that... for now, it feels ascetic and calmly wild.
Here's what it's built with:
- Replicate.com to run ai models via api. They also handle storage of the outputs
- Vercel to host the website
- Nextjs with ChakraUI for the frontend
- Stripe for payments
- Posthog for analytics
- A mix of Vercel and Axiom for monitoring
- A telegram bot to ping my every time someone clicks the buy button (noob enthusiasm ftw)
- Gelato's print-on-demand APIs for getting one-off products made and shipped
It's is at https://selfarama.com/pearl
I'm pretty pleased with it. Simple AF, but with some nice little features... like there's a free shiny thing right away, and like the opengraph images are dynamic, so if anyone shares a link to their own image, their face is on the preview. And it's got almost no buttons!!!
Now I just gotta figure out how to market it, what lessons I'm ready to learn from it, and where to take it next.
Peace!
Your skill is great. But I'm not sure about the product, like who will want his photo with a ponytail or something like this?
You could have make a tool like this, give your photos and provide business ready/professional image from that.
There are thousands way to monetize your this skill.
I want to stay sorta focused on the niche of "regular people who are interested in arty personalized products"
That said, if you or anyone else wants to handle marketing, I would happily build a site that does this for any other nice!! Let's collab
I dont know, man....It sounds like you had an idea which you believed in and had some legs but bailed on it because executing was harder than you hoped....and consoled yourself with an idea you don't believe in and probably doesnt have legs but was easy to execute. Is this a small win to build confidence and get back on the right path, or is this self-defeating behaviour? That is the question only you can answer.
I love the willingness to go dark, and to consider whether it has that quality, but I don't think it's that. It's almost always gonna be "harder than I hoped" because we all have the stars of optimism in our eyes at the start.
But things can actually be too hard for someone. I think that first version was too hard for me, at the time. I've been longing for a new idea that I could execute on with confidence, at least technically, so I could focus on the marketing (which is also a huge blank slate of new skill and learning for me).
I don't agree that "you in a famous painting" prints don't have legs. It was actually something a lot of people asked for when we talked about the books.
I like the way you put it at the end: a small win to build confidence.
It's totally that, but also a small win to practice getting the wins. Focus on the wins. The road to bigness is paved with small wins.
Also... for what it's worth, I haven't bailed on the first idea. I am still processing orders, and have some big plans to use what I've learned from the small win to get back on the right path.
Thanks again for the challenging reply! x