AI Photo & Headshot Generator

Users upload selfies, AI trains a personal model, and they pay for professional-looking photoshoots and headshots.

In plain English

Instead of booking a photographer, people upload a set of selfies and get back polished pictures of themselves. The AI studies those photos so later shots look like the same person, then builds sets such as a work-profile portrait, a dating photo, or a travel-style shoot. Customers pay up front, by pack or by subscription, because preparing each new face takes real computer work. Almost all of the studying, the picture-making, the delivery, and the billing is done by the AI. Pieter Levels still chooses the style packs, checks quality, helps customers, blocks misuse, and lets people delete their photos and the saved face.

Automation95%

Startup costNot disclosed

Time to first $1–3 mo

Monthly range*$100,000 Up to

$0–$100K

DifficultyHard

SourceMRR Ledger — Photo AI founder-reported figures (Pieter Levels) ↗

Operator-Reported

A named, real operator ran this and shared results publicly. The figures are their own.

Not disclosed by the source. Previous figures were estimates, not the operator's number. Each user's training run costs GPU money, so charge before training.

The tool stack

Replicate ↗

Pay per run

Fine-tunes a personal image model (e.g. Flux LoRA) and generates photos.

Stripe ↗

Per-transaction fee

Subscriptions / one-off packs.

Cursor ↗

Free tier; Pro $20/mo

AI-assisted coding of the app.

Prices are the vendors' published list prices as researched in Oct 2026 and change often — always check the vendor's pricing page.

Step-by-step setup

Pick a clear outcome people pay for: LinkedIn headshots, dating-profile photos or travel/influencer shoots. Don't build a generic image tool.
Pipeline: user uploads 10–20 selfies, you fine-tune a LoRA via API, then generate from curated style packs (prompts and presets).
Charge up front (paid packs or subscription) because each training run costs GPU money; show sample results on the pricing page.
Add privacy controls (delete models and photos) and filters against misuse.
Grow by posting your own generated results and building in public on X. Add new style packs regularly to bring users back.

What's automated vs. the human 5%

The machine does

Model training and photo generation
Delivery, billing and upsells

You still do

Curating style packs and quality
Marketing, support and abuse moderation

Overview

Photo AI, a solo project by Pieter Levels (@levelsio), trains a model on a user’s face and then generates photoreal shoots. Levels reported about $105K/month in revenue and about $80K/month in profit in March 2026. His X bio read “PhotoAI.com $86K/m” in October 2026. These are founder-reported figures and not audited through Stripe.

Costs

Startup cost is not disclosed. Each user’s training and generation costs GPU money, so take payment before training.

Realistic income

Levels had a large audience before he launched, and that drove much of his distribution. Copycat headshot apps are everywhere, so a new entrant needs a sharper niche (a profession, a use case or a language) and its own distribution.

Risks

Saturated market. Big platforms add AI photo features too.
Privacy and deepfake misuse, plus payment-processor scrutiny.
GPU cost swings and model licensing terms.

*Income figures are illustrative ranges based on public pricing and typical outcomes reported by practitioners, not guarantees. Most new online businesses earn little or nothing at first. Not financial, legal or tax advice.

Featured in: How to Make Money with Replicate: Eight Documented Paths