AI photo editing credit broker for casual Creative Cloud users

The specific complaint about Firefly credits isn't really about money — it's about surprise. Users who pay for Creative Cloud discover mid-project that Firefly generative credits are a separate meter, not covered by the subscription they assumed included everything. They run out, get blocked, and face a choice: pay again for something they thought they already paid for, or abandon the workflow.

Adobe has a structural incentive to keep credits separate because generative AI inference is genuinely expensive and metered billing lets them shift that cost to power users without raising base subscription prices. The result is a pricing architecture that's rational for Adobe but opaque and frustrating for users — and Adobe has no reason to fix this because it generates incremental revenue from exactly the users who are most engaged.

What's missing is a way for occasional or light Firefly users to access generative editing features without either maintaining a full Creative Cloud subscription or buying Adobe's credit packs at Adobe's price. The gap is specifically: there's no way to buy a small number of generative editing operations at a fair per-use price, through a channel that isn't Adobe itself, with no monthly commitment.

This is a business and not a feature because the underlying dynamic — users needing occasional AI image generation without a subscription — recurs every time someone shares a project, teaches a course, or does a one-off job. The buyer is often not the person paying Adobe; it's students, occasional editors, and small teams where one person holds the subscription and others need a few generations each month.

What to build

Build a web-based generative image editing tool — background replacement, object removal, outpainting, style transfer — that charges per operation with no subscription, where credits are purchased in small packs starting at $3 and never expire.

Where to start

Launch specifically as a tool for educators assigning photo editing projects — they need students to have access to AI editing features for a single semester, and per-student per-semester pricing is far more natural than asking every 19-year-old to maintain a monthly subscription.

The hard part

Inference costs per operation are high enough that low-volume users may not generate enough revenue per customer to cover acquisition cost — the unit economics only work if average credit pack size stays above a threshold that's hard to enforce without annoying users.

How it makes money

Pay-as-you-go credit packs ($3 for 20 operations, $8 for 60, $20 for 200) with no expiry; bulk institutional pricing for educators buying class-wide access by semester.