AI Photo Tool Pricing Explained: Credits vs Subscriptions vs Per-Image

By the ORA Lab team · Updated 26 August 2026 · 9 min read

Two tools sit on your shortlist. One costs $29 a month for 100 credits; the other costs $79 flat with 'unlimited generations'. Which is cheaper? The honest answer, and the reason this article exists, is that the question can't be answered from the pricing page. It depends on what a credit buys, how many generations a usable image really costs you, and what happens to unused capacity at month's end. Vendors aren't necessarily hiding the ball; they're pricing different architectures in the units those architectures make natural. But the effect on buyers is the same: tier names compare apples to weather.

This guide converts every common pricing model into one comparable number, cost per usable image, and flags the fine print that moves that number after you've subscribed. It pairs with our studio-vs-AI cost comparison (which prices AI against traditional shoots) and the 12-test fidelity checklist (because the cheapest tool that fails the count test is the most expensive item on this page).

The four pricing models, decoded

Credits deserve one extra note because they dominate the category: a credit is not an image. In most credit systems, a standard generation costs one credit, but an upscale costs more, a revision costs another generation, and video costs several credits per second. The credit is a metering unit, not a deliverable, read the exchange-rate table, not the headline count.

The multiplier nobody prints: keeper rate

Here is the arithmetic that decides most real-world costs. Suppose a tool charges the equivalent of $0.50 per generation. If one generation in five is listing-ready, a usable image costs $2.50; if four in five are, it costs $0.63. That 4× spread, driven entirely by keeper rate, is wider than the sticker-price spread across most of the market. Keeper rate is a function of the tool's fidelity architecture (a changed product is an automatic discard), its revision behaviour (can you fix one thing without re-rolling everything?), and your own input quality. This is why the cheap-per-credit tool is often the expensive-per-image tool, and why any serious price comparison starts with a free-trial batch on your own SKUs, counting keepers honestly.

The fine print that moves the number

• Credit expiry and rollover: monthly credits that vanish unspent raise your effective rate in every light month, a seasonal catalog (think festive-season spikes) burns cash on the off-season floor.
• What counts as a billable action: revisions, upscales, background removals, and video seconds each consume differently; a workflow with a healthy revision habit can double consumption without doubling output.
• Resolution and export gates: some tools deliver full print-grade resolution only on higher tiers, if your hero images need it, price the tier that actually ships it.
• Licensing and usage rights: confirm commercial use is included at your tier, whether outputs can run in paid ads, and what happens to rights if you cancel. The clause matters more for model-wearing imagery, where likeness terms enter the picture.
• Seat and brand limits: per-seat pricing or single-brand-profile caps quietly multiply the bill for agencies and multi-brand houses.

Price a realistic month, not a tier name

1. 1.Count your monthly SKU flow: new products plus refreshed listings that need imagery. (A 200-SKU catalog refreshing quarterly ≈ 65 to 70 SKUs/month.)
2. 2.Multiply by images per SKU: marketplace main, two to three context or on-model frames, social crops, typically 4 to 6 usable images per SKU.
3. 3.Multiply by attempts per keeper from your own trial batch, not the vendor's implied 1.0. While learning a tool, 2 to 3 is realistic.
4. 4.Convert to each tool's billing unit (credits consumed, generations, delivered images) and read the real monthly price off its actual exchange table.
5. 5.Add the surrounding costs where relevant: operator time per usable image, and any editing pass the tool's exports still need for channel compliance.

Matching model to team

Once prices are comparable, the remaining choice is about how your team works. High-volume, in-house, always-on content operations get the best economics from generous subscriptions or bulk credits, their usage is the predictable kind those models reward, which is the operating mode of velocity-driven D2C brands. Low-volume or seasonal sellers should resist annual commitments entirely and pay per use, eating the higher unit price in exchange for zero idle spend. And teams that want outcomes rather than software, campaign-grade imagery delivered, QA included, nobody learning prompt craft, are better served by per-SKU or service pricing, where the keeper-rate risk sits with the vendor instead of the buyer. That last model is where ORA operates for most clients: you're buying accepted images, not attempts. If you want to see what your realistic month costs on that basis, book a demo with your catalog numbers and we'll price the worksheet with you.