I sold the app I built "just to learn AI". Three weeks later I launched its opposite.

Hey IH,

Last year I posted on Reddit about ArtScan, an iOS app I built in summer 2024 just to learn the OpenAI API. Point your camera at a painting, get the story behind it. No market research, no launch, no plan. It somehow grew to ~20k downloads and ~$10k annual revenue on 30 to 50 organic installs a day, ranking for keywords I never targeted. I spent a year not fully understanding why it worked.

In July, I sold it on Flippa. The fun detail: the buyer had already bought two of my smaller apps before. Repeat customers exist in the micro-acquisition world too, apparently. It was a five-figure exit (low-to-mid, and I will leave it there). Not life-changing money, but very real money for a side project born out of curiosity.

Most people would take the win and rest. I took the win and immediately built the opposite product.

From "what painting is this?" to "make me the painting"

ArtScan answered an identification question. Curious people, museum visitors, folks finding canvases in their attic. Lovely audience, low willingness to pay: my average user was worth cents.

My new project, Ninth Salon, flips the direction: you upload a photo and it turns you (or your kid, or your grandma, or your backyard) into a classical oil portrait. Renaissance, Baroque, Belle Époque. It started as a real gift: I turned a photo of our daughter into a Renaissance portrait for my husband, and watching his reaction taught me more about intent than a year of ArtScan analytics. Identification is curiosity. A portrait of someone you love is a gift, and gifts have budgets.

What I kept from ArtScan

• The niche. Art and AI is where I like spending my nights. ( I am myself an art lover ^^)
• The stack philosophy: boring tech I already know. The frontend is vanilla JS, no framework, 43 KB gzipped.
• Shipping fast: three days from first commit to first real sales.
• Staying solo, with a full-time job, building at night.

What I changed on purpose

• Web instead of App Store. No 30% cut, no review queue, and I own the relationship. ArtScan lived and died by keywords I could not see or control.
• The pricing model is the product. Anyone can generate a portrait for free, no signup, and see the full result with a watermark. You only pay to unlock the clean file. I charge at the emotional peak, not at the door. Conversion so far says the bet is right.
• A physical product on top. The portrait can be printed on gallery canvas and delivered ready to hang. One canvas order has the margin of many credit packs. Print on demand means my manual work per order is near zero.
• Three languages from day one. English, French, Spanish. The French and Spanish SERPs are dramatically softer than English for the same intent.
• SEO on purpose this time. ArtScan ranked by accident. This time I shipped 22 blog articles in the first week, aimed at gift intent, before Google even knew the domain existed.
• And the loop I enjoy most: the ArtScan money is funding the ads. My old app is literally buying customers for my new one.

Where it is today

Three weeks in: first real sales (tiny, but strangers paying), canvas pipeline live, and a launch scheduled on Uneed next week. I will report numbers as they become interesting.

The decision I keep second-guessing (genuinely asking):

There is no pricing anywhere on my landing page. Anonymous visitors see zero prices, zero "free", zero paywall talk until they have generated their portrait and love it. Classic CRO says I am burning trust. My theory is that for an emotional purchase, showing the price before the wow moment only gives people a reason to leave. Am I wrong?

Happy to answer anything about the sale, the pricing model, or building AI wrappers that people actually pay for.

PS: still have a full-time job, still building at night.
PSS: if you want to see the new one: https://ninthsalon.com
PSSS: still a lady, BTW. "Bros" still aren't the only ones who code ;)