Capcom has announced that RE Engine, the tech behind Resident Evil 4’s remake, Street Fighter 6, Dragon’s Dogma 2 and Monster Hunter Wilds, is being gradually rebuilt into what the company calls an “AI-generation game engine.” IGN’s write-up carries the company line that its goal is “a future where we create games together with AI.” Eurogamer connects the plan to Capcom’s next-generation engine work, the REX project.
That phrase is doing a lot of work, and PC Gamer is right to flag it. An “AI-generation” engine and a “generative AI” engine are not the same claim. One is an engine built for the current era of development, with machine learning woven through the toolchain. The other is a machine that hallucinates a game world on demand. Capcom’s wording points at the first. The headlines will get read as the second.
What this probably actually means
The boring version is the likely one: asset pipelines. Automated LODs, retopology, upscaled or inferred textures, animation retargeting, lipsync for a dozen localisations, procedural variation on rocks and crates and NPC faces, QA bots grinding a level for collision gaps overnight. None of that is futurism. Ubisoft, EA and Square Enix are all somewhere along that road already, and middleware vendors have been selling pieces of it for years. Capcom is unusual mainly in saying it out loud as an engine-level commitment rather than a quiet internal tools memo.
The version worth being suspicious of is any suggestion that generated content reaches the player unexamined. Capcom’s whole competitive edge is handcraft. The reason the RE4 remake lands is that someone agonised over the exact timing of a parry. You cannot infer that from a training set.
The problem RE Engine actually has
Here is my objection to the framing, though. RE Engine’s bottleneck right now is not that artists make textures too slowly. It is that recent Capcom releases ship with PC performance that ranges from rough to indefensible. Monster Hunter Wilds was the clearest case, a game whose own publisher ended up acknowledging the complaints. Dragon’s Dogma 2 had its own CPU wall in the city. Those are scheduling and optimisation failures as much as engine failures, and a smarter asset pipeline does not fix either. If anything, tooling that makes it cheaper to produce more stuff raises the odds of shipping more stuff that runs badly.
So the pitch I want to hear from Capcom is not about co-creating with AI. It is about frame pacing, shader compilation and memory budgets on a five-year-old console generation.
There is also the thing nobody in these announcements addresses: whose work the models learned from. Capcom has decades of internal art to train on, which puts it in a far cleaner position than most. It has not said that is what it is doing.
No timeline has been given, no products named, and no detail on which parts of the pipeline come first.