Only Anime News Network has carried this one in English so far, so the second source below covers the AI-animation strand rather than the name change itself.
Read the paperwork rather than the name. Topco Media — the Korean publisher behind the webtoon platform Toptoon and the web-novel platform Novelpia — approved a change to Ankey AX Co., Ltd. at an extraordinary shareholders’ meeting on 10 September 2026. At the same time it added AI content, AI chatbots, virtual humans and AI solutions to its stated business purposes.
Filing virtual humans as a business purpose is a statement about where the revenue is expected to come from. Faster comics does not need that clause.
The stack they already built
The rename lands on top of about a year of work. In September 2025 Topco announced short-form animation for its webtoons, driven in-house: motion over existing panels, with AI-produced shorts as a premium-tier perk. By early December 2025 it was reporting that its first AI-produced adult animation, Office Romances Are Strictly Forbidden, had done well enough in Taiwan to justify a global rollout, and it put a number on it — returns of roughly five times the original webtoon’s production cost. Treat that figure as the company’s own, because it is; nobody has audited it.
The pipeline description is harder to wave away. An in-house AI studio, expanded NVIDIA GPU capacity, and a custom production chain combining Grok, NovelAI and ComfyUI, producing output the company says runs three to five times faster than its traditional process.
ComfyUI sitting in the middle of that is the detail I keep turning over. It is a node graph. Load a checkpoint, mask a region, run a controlnet off a pose or a depth pass, composite, save. You can version it, hand it to someone else, and diff it when a shot comes back wrong. That is a pipeline tool in the sense a studio means the word, and it is a long way from the “type a sentence, get a picture” framing most coverage still uses. Somebody there has built graphs.
Why webtoons went first
Every format that generative video has chewed through early had the same property: the source art was already separated.
Vertical-scroll comics are the extreme case. A webtoon episode is produced as stacked layers — line, flats, shading, background, effects, lettering — because it has to ship weekly and the work is split across a team. Character sheets are locked early and enforced, since three different colourists have to produce the same face. Backgrounds are frequently blocked in 3D and traced or rendered over, which means the geometry exists somewhere in the studio. There is no camera solve, no lighting continuity between shots, no lip-sync tradition to violate.
Hand that to an image-to-video model and you are asking it for something close to its easy case: hold this design, move this layer, give me four seconds.
Which is the part worth sitting with. The division of labour that made webtoon studios fast — every stage handing the next a clean, standardised file — is the same division that makes a stage removable. An assembly line is legible. A legible process is one you can describe to a machine.
The number that does not mean what it says
Three to five times faster output is a real claim and I believe some version of it. Animation schedules are governed by revisions, and drawing frames sits well down the list of what consumes them.
Notes are the expensive thing. A director looks at a cut, says the hand reads as wrong, and somebody fixes that hand while the other four hundred frames stay exactly as they were. Diffusion does not do that. You regenerate and take a new roll of the dice on everything else in the shot, or you paint over the output by hand, at which point the human labour you removed walks back in through the compositing door. Level-5 ran into the shape of this problem from the other direction when Akihiro Hino apologised for AI and restated the five-to-two plan in the same broadcast — the ambition is schedule compression, and schedules are made of revisions.
So a five-times throughput figure is most plausible on work nobody will give notes on. Short-form promotional loops. Adult content sold on volume. Chat avatars. Look at the business purposes again and that is roughly the product list.
What this costs the people drawing
I have watched enough pipeline tooling arrive to know the pattern, and the honest version is not that the tool is bad. The places where generative work earns its keep in a production pipeline are small, unglamorous and compounding — cleanup, variants, the fourth iteration of something already approved. A background plate you needed anyway, in a style already locked.
What changed here is the size of the unit being replaced. A whole stage of the line, rather than a task inside somebody’s day.
Toptoon’s catalogue was drawn by working artists under contract, and the company’s plan now is to develop those properties into AI-generated conversations, stories and images. Whether the original creators see anything from a derivative their contract never imagined depends entirely on Korean platform contracts nobody outside those studios has read, and I am not going to guess at terms I cannot see. It is the question I would want answered before signing the next one.
Ankey AX also shipped Toptoon Chat, a service that lets readers talk one-to-one with characters from its webtoons, first in Korea and since extended to Japan, Taiwan and North America. A character someone designed, now answering questions the designer never wrote. That is the business, described plainly, and it is why the new name has AX in it rather than toon.