Photo AI logo
Photo AI
05 Jul 2026 · 4 min read
AVOID

The $138k a month app you should not copy

A one man AI photo studio pulling $1.6M a year, and the wall standing where the door used to be.

MRR
$138k
Team
1
Price
$19
Started
2023
AI · Photos · Consumer
Saturation

This is the screenshot that makes builders quit their jobs. One guy, zero employees, an app that turns your selfies into studio photos, pulling around $138,000 a month. It is the most copied dream in indie software right now, and this is the first teardown where I am going to talk you out of it.

The product itself is simple to describe. You upload a handful of selfies, Photo AI trains a model of your face, and out comes a studio's worth of photos of you, the LinkedIn headshot, the dating profile, you on a mountain you have never climbed. It launched in February 2023, did $5,400 in its first week, hit $132k a month inside eighteen months, and sits around $138k a month today at prices starting from $19.indiehackers.com+2

How it grew is the part most people skip past, and it is the part that decides everything. Pieter Levels built it in front of roughly 600,000 followers he had spent a decade earning through Nomad List and Remote OK. When he ships, launch day looks like a stadium. When you ship, launch day looks like your group chat. Same product, entirely different business.fast-saas.com

And the app itself is not the hard part, because the heavy lifting is rented. The face training and the image generation run on model APIs anyone can call, so the actual build is a weekend of gluing a fine-tune endpoint to a checkout page. Which sounds like great news for you, until you realize everyone else heard it too.

Because they came by the busload. Reviewers now test thirty five plus headshot tools to write their comparison posts, the space is packed with white label clones of clones, and the price war is already at the floor, $27 here, $29 there. HeadshotPro, the biggest pure headshot player, read the room and fled upmarket into corporate team accounts. All of them are fighting over a customer who, and this is the quiet killer, needs the product roughly once a year.medium.com+2

Then the ground itself moved. ChatGPT now generates photoreal images of you natively, about 200 a day on its twenty dollar plan, and the free tier does it too. The entire category is turning into a prompt inside a general tool that a billion people already have installed. That is not a competitor, that is the tide coming in.aivideobootcamp.com

THE GAP

Everyone stares at this number and sees the same opening, make it cheaper. That opening closed two years ago, and what is standing there now is a wall.

Walk through the logic the way every would-be cloner does, because it feels airtight. The product is a weekend of work, the revenue is public, the market is proven, therefore build the cheaper one and take a slice. The flaw is that reasoning this obvious gets acted on at industrial scale. When the build is trivial and the number is on a public dashboard, the market fills before you finish reading the tutorial, and the fifty clones already fighting at $27 are the proof. When code is a weekend, code is not the business, and a cheaper version of a weekend is worth exactly nothing.

Now look at the customer, because the math is worse than the crowd. A headshot is a once a year purchase at best, dressed up in a subscription it will cancel next month. There is no compounding here, no growing account, no switching cost. Every single month, every player in this market starts back at zero and buys its customers again, which is why the funded clones with ad budgets are the only ones who can stomach it. You would not be buying recurring revenue, you would be stepping onto a treadmill that eats margin.

And the thing you actually envy, the $138k, was never really the app. Photo AI's moat is the 600,000 people who watch Levels build. He can launch a landing page and outdraw your entire year of marketing before lunch. Clone the product and you have cloned the one part that was never scarce. The distribution is the company, and you cannot fork a following.fast-saas.com

Even the king is exposed, which tells you how closed this room really is. When ChatGPT hands out photoreal self portraits on a plan people already pay for, every wrapper in the category is living on borrowed time, and Levels is reportedly earning seventy percent of his income from this one product while that tide rises. If the strongest player in the market is this exposed, the fifty-first entrant is not a business plan, it is a donation.ppc.land

So the gap this teardown hands you is a lens, not a wedge. Public revenue is a clone magnet, the screenshot you envy is the same screenshot fifty other builders saw eighteen months before you, and by the time a number goes viral the opening behind it is gone. The real openings sit where revenue hides, in the boring vertical software nobody screenshots, sold to people who do not read launch threads. That is the hunting ground. This market is a museum, walk through it, learn from it, and build somewhere else.

VerdictAVOIDThe opening closed before you found it. Keep the lesson, keep your months.

The build stack

Pro

You could stand this up in a weekend for well under a hundred a month, and that is exactly the warning. A stack this cheap is a market this crowded.

Next.js on VercelFree tier, then ~$20/mo

The app, the checkout, the gallery.

Cheaper:Any $5/mo VPS, though it hardly matters here.

Replicate or FALUsage based, a few dollars per trained customer

Rent the face fine-tuning and image generation models.

Cheaper:Open weights on a rented GPU, more ops for less margin bleed.

Cloudflare R2~$0.015/GB, cents early on

Store the uploads and the generated photos.

Cheaper:Already the cheap option.

StripeNo monthly fee, 2.9% + 30c per charge

Subscriptions the customer cancels after one use.

Cheaper:Lemon Squeezy as merchant of record, ~5% per sale.

ResendFree 3k/mo, then $20/mo

Photos-are-ready emails.

Cheaper:Amazon SES at pennies per thousand.

Prices are public list prices and change over time. Treat them as a starting estimate, not a quote.

The build stack is for members

Every teardown comes with the exact tools to replicate it, their monthly cost, and cheaper swaps. The whole library for $12/mo.

Unlock the stack

The full clone playbook

Pro

Step by step. The build and the go-to-market.

  1. 01
    Count the clones before you code

    Search 'best [product] tools' before you build anything. If reviewers are ranking thirty five of them to fill a listicle, the listicle is the verdict. The market voted before you arrived, and you do not out-ship a stampede.

  2. 02
    Check what the purchase really is

    Ask how often one customer genuinely needs this. A once a year need wearing a monthly subscription is churn in a costume, and a churn business is an ads business, and an ads business belongs to whoever has the deepest pockets. That is never the solo builder.

  3. 03
    Find the real moat before you covet the number

    If the founder's audience is the distribution, the product is a storefront on land you do not own. Cloning the store without the foot traffic gets you an empty shop. You cannot fork a following, so price the audience into every public number you envy.

  4. 04
    Watch what the platforms are absorbing

    Before entering any AI wrapper market, spend ten minutes doing the job with the free general tools. If ChatGPT gets you eighty percent of the way there today, it gets a hundred next year, and every dedicated app in that lane is on borrowed time.

  5. 05
    Aim the same month somewhere real

    The energy this clone would have burned is enough to ship a boring vertical tool for a trade that has never read a launch thread. Revenue that hides is revenue nobody is swarming. Go where there are no screenshots, that is where the openings still live.

The full playbook is for members

Every gap comes with the step-by-step build and go-to-market. The whole searchable library for $12/mo.

Unlock the playbooks