Capcom is rebuilding the RE Engine as an "AI-generation engine"

Capcom has announced a multi-phase project to adapt its proprietary RE Engine into an "AI-generation engine." The engine was built in 2014 for Resident Evil 7: Biohazard, which launched in 2017, and it's estimated that more than 2,000 developers use it.

Programmer Satoshi Ishida laid it out at Capcom's annual tech conference, in a talk titled "The Outlook and Future of the REX Project, Further Evolving the RE Engine for the Next Generation."

The pieces

• RE:Dox — standardised data processing
• RE:UI — an accelerated development toolkit
• RE:Log — ID-based logging
• RE:Flows — visual scripting
• RE:Runtime — a reworked runtime core
• RE:C++ — a refined execution framework

Ishida said integrating AI is aimed at assisting development, generating and interpreting code, automating playtesting for debugging and optimising performance — moving toward a future where developers "create games together with AI." Some improvements will be open-sourced to support AI training.

Read it alongside what Capcom said before

Earlier this year Capcom emphasised it would not "implement any generative AI assets" into its games, and would instead use the technology to "improve efficiency and productivity."

That distinction is the whole story here. Capcom is not proposing AI-drawn monsters. It is proposing AI that writes and reads code, runs your playtest for you, and finds the frame that drops. The company also recently had Resident Evil Requiem appear in a demo of Nvidia's DLSS 5, which drew criticism for using AI to enhance character models — including protagonist Grace Ashcroft.

The line being drawn

Compare the three positions from the same month:

• Epic reportedly told a team: no more hand-written code, all art from full AI workflows.
• Sega said it won't "entrust the creative aspects of games to AI," while using it for publishing and admin efficiency.
• Capcom wants AI inside the engine and the pipeline, and is explicit that finished assets stay human.

Three different companies, three genuinely different answers. The industry hasn't converged on whether AI belongs in games. It has started converging on a quieter question: how much of the invisible work is a model allowed to do.