Layer

Here is a fact about making games that does not make it into the trailers: a modern studio consumes an enormous quantity of art. Every live-ops event needs its own banner set. Every user-acquisition campaign burns through ad variants the way a car burns through fuel on a highway. Every seasonal skin, every in-game icon, every localized store screenshot is a small production job, and there are thousands of them, and they are all due Friday. The creativity is not the bottleneck. The photocopier is.

Layer, a San Francisco company founded in 2023, decided this was the interesting problem. Not "can AI make a pretty picture" - the internet had that covered - but "can AI make three hundred on-brand pretty pictures that fit inside a studio's existing pipeline without an artist wanting to throw the laptop out the window." Those are very different questions, and the second one is worth more money.

The artist-first loop

The mechanism Layer built is less a text box and more a loop. A studio trains a model on its own art. The model generates assets in that house style. The artist curates, edits and refines them on a real-time canvas. The good outputs feed back into the model. Repeat. The point of this loop is that the artist never leaves the driver's seat - the AI does the volume, the human keeps the taste - which happens to be exactly the pitch you need to sell into a AAA studio that is protective of its brand.

This is the part worth pausing on, because it is where the "AI will replace artists" discourse tends to go sideways. Layer's product spec is the opposite premise. If the model replaced the artist, the house style would drift, the brand would blur, and the studio would notice by the second campaign. So the artist stays, and the tool's job is to make that one artist as productive as a small team. Studios say the result is a two-to-eight-times bump in production speed. That is a range wide enough to be honest about - the number depends entirely on what you are making.

Model-agnostic, on purpose

Layer describes itself as model-agnostic, which sounds like a dull architecture footnote and is actually the whole strategy. The company deliberately refuses to marry a single AI model. The reasoning is straightforward: no model wins forever, the frontier moves every few months, and a studio's production pipeline needs to outlive the current hype cycle. By staying model-agnostic, Layer lets the pipeline survive the churn underneath it. Boring decisions, made on purpose, tend to be the load-bearing ones.

The wedge, and the widening of it

Layer did not try to boil the ocean. It started with 2D image generation, proved it worked, and only then expanded into 3D, then video, then audio. The integrations tell the same story of meeting artists where they already are: Photoshop, Illustrator, Blender, Maya, Unreal, Unity, Godot, plus API and webhook access for the teams that want to wire it into their own tooling. Assets export straight into the game engine. The friction the company removed is the friction of moving work between tools, which is the friction that actually eats a production schedule.

The people behind it have a slightly unusual shape for a creative-tools startup. CEO Volkan Gurel studied at MIT under Marvin Minsky and Pat Winston - two of the field's founding figures - and then spent years on scalable infrastructure at Coinbase and AI work at Airware. That is an infrastructure engineer's résumé, not an art director's, and it shows in the product: Layer reads less like a toy and more like plumbing. Co-founder Burcu Hakguder, the chief revenue officer, came out of Unity Technologies, which is roughly the ideal place to learn what game studios will and will not pay for.

Who is buying

The customer list is the argument. Zynga, SciPlay, Tripledot Studios and Huuuge Games are among the 300-plus studios using Layer, alongside more than ten thousand artists, producers and growth teams. These are not hobbyists experimenting with a free tier. They are performance-marketing and live-ops operations with real deadlines and real brand guidelines, which is the exact audience that separates a novelty from an enterprise tool. The unglamorous features - SOC 2 compliance, role-based access, collaborative workspaces - exist precisely because that audience asks for them before signing.

In May 2025 Layer closed a $6.5 million seed round, on top of the $1.8 million it raised in late 2023, bringing the total past $8 million. The round was led by Arcadia, the fund of Akin Babayigit, with an investment vehicle connected to King co-founder Riccardo Zacconi and existing backer e2.vc. When the person who helped build Candy Crush into a mobile empire writes a check for your game-production tools, it is at minimum a signal that you have read the industry correctly. Whether Layer becomes the default creative infrastructure for games is still an open question. But it has picked a real problem, built a boring and useful answer to it, and gotten the people who feel that problem most acutely to pay for the fix. That is usually how these things start.