x-octo home Business judgment on AI products
中文

Real-revenue cases · AI

RePho: Restore Old Photos

Nice-to-have

An iOS app that uses AI to restore old photos — repairing scratches and tears, colorizing black-and-white images, sharpening blur, removing unwanted objects, and turning stills into short video clips — sold on a subscription.

Stripe-verified App Store 自然搜索素材未提及
Monthly revenue
$187 / mo
Primary source
View original

01

Real-demand verdict

Real-demand verdict

Nice-to-have

Restoring old photos is a genuine pre-existing need once handled by manual retouching or Photoshop, and AI turns it into a one-tap action, so some people do pay. But at $187/month the paying base is tiny and looks closer to one-off impulse use than a durable subscription.

How it makes money

iOS users pay a subscription; Stripe-verified monthly revenue is $187. The source does not disclose pricing tiers.

What old behavior it replaces

Manual retouching by a photo editor, or doing it yourself in Photoshop

02

Where the first customers came from

Not mentioned in the source

Acquisition channels App Store 自然搜索素材未提及

03

Tactics you can copy

  1. 01Bundle restoration, colorization, sharpening, object removal and animation into one subscription so the feature count justifies the price
  2. 02Pick an emotionally charged, visually obvious category like old-photo repair, since the output is inherently shareable
  3. 03Start with an iOS subscription and Stripe-verified revenue — prove payment works before chasing scale
Moving it to an AI business

For AI builders: the 'upload old asset, get a visibly better output' shape transfers directly to old-video repair, audio denoising, or handwriting digitization. The key move is bundling several single-purpose AI capabilities into one subscription rather than charging per use.

04

Evidence and limits

Self-reported numbers are unaudited — treat them as leads, not facts
Stripe-verified

Revenue basis · Stripe-verified (TrustMRR)

What evidence is missing

At $187/month there is no retention or churn data, so the 'want' verdict rests on revenue scale, not on renewal behavior
No pricing, tiers, or paying-user count, so unit economics can't be checked
Acquisition channels and cold-start method are entirely absent, so replicability is unverified