Layer

Where Layer sits

Layer (layer.ai) calls itself "the AI Operating System for Creative Teams," and after a full pass through the product site, pricing, help center and customer stories, I think the operating-system framing is closer to the truth than the usual "AI art tool" label. This is a production workspace aimed at game studios and entertainment brands: one place where artists, UA managers and producers generate, edit and ship images, video, 3D and audio while the studio keeps its own visual identity. The company was founded by Mehmet Ecevit, who sold Gram Games to Zynga in 2018, and the product design clearly comes from someone who watched art become the bottleneck in mobile game production. That origin shows up in details competitors skip: PSD export for Photoshop finishing, reference sets trained on a studio's own approved art, per-title workspaces with separate budgets, and billing that follows output rather than headcount.

What I look for in this category is whether a platform survives contact with a real production pipeline. Anyone can bolt a text box to Stable Diffusion. Fewer tools can keep a live-ops calendar in style for three years without the art director quietly going back to hand-drawing the important pieces. Layer's bet is that consistency, workflow automation and enterprise controls are the product, and the 300-plus models are interchangeable parts behind it.

What Layer actually does

The core surface is a shared workspace with a pooled currency called Creative Units (CUs). Every generation across every modality draws from the same balance, and there are no per-seat charges — you can add the whole studio to a workspace and the bill doesn't move. Four modalities are covered: image, video, 3D and audio. My read of the model catalog page during testing: 103 image models, 136 video models, 38 3D models, 33 audio models and 32 task-specific creative tools, for 310 models in total — although the same company's meta description later in my session advertised "399+ models," so treat the catalog as a moving target. Models come from Google, OpenAI, Black Forest Labs, Kling, Runway and others, and new ones land in the same workspace, so a workflow can swap one node for a better model without being rebuilt.

The consistency story rests on custom style training. You upload a library of approved art and Layer trains a reference set on it; generations then draw on that set across character skins, UI elements, icons and marketing creative. The workspace supports batch editing and refinement (not just one-shot generation), a node-based workflow builder that producers can run like an app — type the event, press Run — and asset exports into Unity, Unreal and, per the LBC Studios case study, layered PSDs for Photoshop.

On the enterprise side, Layer is SOC 2 Type II certified (the company positions itself as the first gaming-focused generative AI platform to reach that bar), with SSO across Google, Microsoft, Okta, Azure AD and any SAML 2.0 provider, SCIM provisioning, role-based access, audit logs, EU data residency and a stated policy of never training foundation models on customer work. There's a documented REST API plus MCP server entry points for agent-driven use.

My testing notes

Two things I checked carefully, because pricing churn is where these tools usually disappoint.

First, the pricing model itself. Since September 9, 2025, Layer charges $0 per seat and bills purely in Creative Units. The help center's numbers at time of writing: free accounts get 50 CUs to try the platform (the announcement post says a business email adds 250 more; the help center says 300 — a small but real inconsistency I couldn't reconcile), paid subscriptions start at $10 for 300 units with a 50% first-purchase discount on 300/500/2,000-unit packs, ongoing 1,000-unit packages run $60/month ($0.06 per unit), and one-off packs are $40 per 500 units ($0.08 per unit). The pricing page includes a calculator — sliders for roughly up to 6,000 images, 1,000 video seconds, 300 3D models or 6,000 audio hours — with the honest caveat that estimates "may vary based on usage," because different models burn units at different rates. A billing dashboard shows burn rate per action and model before changes take effect. For a studio generating hundreds of assets a day, this is meaningfully cheaper than per-seat licensing; for a solo dev generating ten images a month it is neither the cheapest nor the simplest option.

Second, the productivity claims. The LBC Studios case study (a 40-person Vancouver mobile studio) is the most concrete evidence published: about 10 hand-drawn characters were used to train a reference set, the art lead reports the model picked up "fine details such as hand proportions and the small line breaks in our outline strokes," and the studio shipped 10 style-consistent characters plus vehicles, icons and props in a work week — roughly 40 hours of work versus an estimated 320 hours traditionally, an 8x production increase and 250+ hours saved. That is a vendor-published case study with named quotes, not my own measurement, but the specificity (hours per asset class, PSD finishing in Photoshop, curating AI variations before commit) reads like a real workflow rather than marketing air. The mobile-games page claims 400+ studios on the platform — Tripledot, Zynga, King, Peak Games, Socialpoint and others are named — and 3–5x output increases across case studies. I treat both as vendor claims.

One more observation from my walkthrough: the security posture is unusually concrete for this niche. "No training on your work," EU storage, workspace isolation and audit logs are exactly the asks a studio with an unreleased title makes, and Layer publishes them without a sales call (the SOC 2 report is available on request).

What it does not do

Layer is a generation and workflow layer, not a DCC tool. Video and 3D outputs are production-ready in the marketing sense; in practice they are excellent drafts. The LBC workflow explicitly ends in Photoshop for final touches, and 3D model generation (38 models) will not hand you a rigged, LOD'd asset ready for Unity — it gives you a strong starting mesh or texture set. Style training also isn't zero-setup: LBC needed a curated seed library of about 10 approved pieces, and the case study itself shows humans scoring and selecting AI variations. Budgeting is consumption-based, which means cost predictability depends on your own discipline; heavy video use can burn units far faster than image work, and only the billing dashboard tells you. The free plan, post-September 2025, is a tasting menu — a one-time allocation of units rather than ongoing free generation. And because Layer aggregates third-party models, output quality shifts whenever an upstream model is retired or repriced; your workflow survives (that's the point of the node design), but your exact results won't be identical month to month. Finally, the inconsistent model count across their own pages (310 in the catalog breakdown, 399+ in one meta description, "300+" elsewhere) is a small honesty wart I'd like to see cleaned up.

Verdict

Layer is the most production-minded generative platform I've reviewed for game and entertainment teams. The combination of style-consistent reference sets, node workflows producers can run without an artist, four modalities on one balance, and real enterprise security is what a scaling studio actually needs, and the $0-per-seat model removes the usual collaboration tax. It is rated 4.2 on the rubric — core functionality depth carries it (40% of the score), with strong value for high-volume teams, a usability curve that rewards committed studios rather than casual users, a maturing developer ecosystem, and solid reliability signals from SOC 2 and named customers. If you produce creative volume and your problem is consistency plus throughput, Layer belongs on your shortlist. If you want the occasional avatar or banner, a simpler single-model tool will cost you less attention.

Sources

• https://www.layer.ai/
• https://layer.ai/pricing
• https://help.layer.ai/en/articles/12037310-guide-to-layer-s-subscriptions-pricing
• https://layer.ai/blog/pricing-updates-2025
• https://layer.ai/success-stories/lbc-studios
• https://layer.ai/industries/mobile-games
• https://layer.ai/blog/unity-layer-student-game-creators
• https://layer.ai/glossary/generative-ai-for-games
• https://wavel.io/ai-tools/layer-ai
• https://rightaichoice.com/tools/layer-ai