Identity-locked AI vs traditional product shoots: a cost & quality breakdown

The honest cost-and-quality comparison between AI product photography and a traditional studio shoot has changed in the last 18 months, and the variable that moved is identity-lock. This post walks through what identity-lock means in practice, why it changed the quality bar, and how the per-SKU economics actually compare in 2026.

What "identity-lock" really means

Identity-lock is the property that the same model face — and the same product — show up consistently across every render in a shot plan, across batches, across drops. It's the difference between an AI photoshoot and a folder of inconsistent renders.

The technical bar today, on Kreton, looks like this:

• ArcFace face-embedding verification. Every rendered face is compared to the locked reference face. Cosine distance above 0.35 is a fail.
• Inpaint-on-drift. A failing shot doesn't regenerate from scratch — only the drifted region is inpainted. This is both faster and produces less variance.
• SIFT structural match. The rendered product is compared against the source product photo for colour, texture and hardware. Failures are routed to the same inpaint loop.
• Per-asset provenance. Each render carries metadata describing how it was produced — useful for both internal QA and jurisdictional AI-imagery disclosure rules.

Why this changes the quality bar versus a studio shoot

Studio photography has one thing AI doesn't, traditionally: a fixed real subject. The same model, the same product. But it also has drift — lighting, fatigue, day-to-day model variance, retouch inconsistencies between artists.

Identity-locked AI inverts the trade-off. The model face is algorithmically fixed across every render (no shoot-day fatigue), and the product is structurally pinned to the source photo (no between-shot drift). What it loses is the photographer's eye — composition decisions made in the moment.

Net effect: for repeatable PDP imagery (the bulk of what a catalog needs), the AI bar is now higher than what a typical 2-day shoot delivers. For editorial and campaign work, where the photographer's eye is the point, the studio still wins.

The per-SKU economics, side by side

Honest numbers, end-to-end including retouch, retries and merchandising time.

Studio shoot, 8-shot PDP plan + lifestyle clip

• $80–$400 per SKU end-to-end.
• 2–10 day turnaround.
• Brand consistency depends on model availability and photographer continuity.
• Editorial output is excellent.

Identity-locked AI, same 8-shot plan

• $3–$6 per SKU end-to-end on Kreton's pay-as-you-go.
• ~90 seconds generation; minutes to approve and publish.
• Brand consistency baked in via identity-lock.
• Editorial output is good but not the same as a 10-year veteran fashion photographer.

Where AI still doesn't win

Three categories where the studio remains the right call:

• Brand films. Long-form video where art-direction and the photographer's eye are the point.
• Campaign hero shots. The once-a-quarter image that defines a brand season. AI gets close but a studio shoot gives you the edge cases — the unexpected moment, the art-direction choice — that defines a campaign.
• Markets that explicitly require human-shot imagery for PDP. Rare, but a few categories and jurisdictions still require it. Provenance metadata makes this easy to comply with on either side.

What this looks like operationally

Teams that have switched typically keep ~10% of their photography budget on the studio for editorial / campaign work, and route the 90% bulk-PDP volume to identity-locked AI. The studio team becomes the curators of the brand's pose library, lighting profile and model face — which then drives the AI shot plan.

Want to see the per-SKU economics on your own catalog?

Start free with 20 renders, no credit card. Or check the broader catalog playbook if you're sizing this up across a whole drop.