Recompilation Developers Push Back as AI-Made N64 Fan Ports Spark "Slopcomp" Debate

A growing dispute inside the fan-driven Nintendo 64 recompilation community has put AI-assisted projects under a harsh spotlight. Developers behind established, human-led efforts to bring classic games to modern PCs are criticizing what they call "slopcomps": recompilation projects reportedly built with generative AI and presented alongside more conventional reverse-engineering work.

The immediate flashpoint is a project called Banjo-Tooie Recompiled, which appeared online in late September. The project prompted criticism from Darío, also known online as DarioSamo, a software developer involved with the earlier unofficial PC recompilation of the first Banjo-Kazooie. His objection was not simply that another group was tackling Rare's Nintendo 64 sequel. Instead, he argued that the new effort's AI-generated code, technical approach, and branding risk damaging a scene built on years of painstaking human research.

Recompilation projects have become one of the most interesting corners of retro game preservation and PC modding in recent years. Rather than merely emulating original console hardware, recompilation tools help translate a game's original program logic into software that can run natively on current systems. The results can enable features that are difficult or impossible to achieve through standard emulation alone, including modern rendering options, ultrawide support, high frame rates, controller improvements, and extensive modding possibilities.

Embedded X/Twitter post

That work has also fueled unofficial PC versions of several beloved Nintendo 64 games. These projects generally require players to provide their own legitimate game files, while the community-created tools and patches reconstruct the game for modern hardware. They exist in a legally and technically complicated space, but their appeal is clear: they can make decades-old games feel dramatically more accessible without abandoning their original design.

For useful background on this topic, read Capcom Details How AI Could Assist RE Engine Development Without Replacing Creators.

Human-led projects draw a line on generative AI

On October 1, Darío sought to clearly separate the teams behind Zelda: Recompiled and Banjo: Recompiled from the new Banjo-Tooie project. He said the latter is not connected to their work and described it as entirely AI-generated. He also said the existing teams have an explicit policy against using AI in the development of their recompilation projects.

For Darío, the issue is not merely ideological. He argued that a polished recompilation requires careful research into the original game, a robust framework, and purpose-built rendering technology. Treating that work as a starting point for an AI-generated rush toward another port, he said, disrespects the extensive effort already invested by human contributors.

He further emphasized that the group had already expressed an interest in eventually creating a human-made Banjo-Tooie recompilation. That context has made the shared naming especially contentious. Although "recompiled" is a broad technical description rather than a proprietary label, similarly titled projects can make it harder for players to tell which releases are connected, what methods were used to make them, and whether they meet the community's expectations for testing and maintainability.

Darío called for the AI-assisted project to adopt another name and criticized its implementation. In particular, he said it does not work with the patching approach normally expected by the recompilation community. His concern is that patching recompiled code and memory from what he considers the wrong side can make future mod development difficult or unmanageable.

Why modding is central to the argument

To players, the distinction between two fan-made PC ports may initially sound academic. If a game launches and looks sharper on a modern display, it can be tempting to judge the project solely on whether it appears functional. Recompilation developers, however, are stressing that a release's long-term architecture matters as much as an early video demonstration.

Modders need understandable, stable, and maintainable systems. A recompilation project can become a foundation for texture packs, gameplay adjustments, randomizers, accessibility features, bug fixes, new camera options, and fan-made expansions. That work depends on contributors being able to identify how a port operates, track changes over time, and reliably apply patches without breaking unrelated elements.

The criticism of AI-assisted recompilations is therefore partly a criticism of technical debt. Code can appear to work while still being fragile, opaque, or difficult to extend. If generated code has not been closely reviewed and tested by people who understand the original game and the recompilation pipeline, developers worry that subtle problems may not emerge until someone attempts more ambitious modifications.

There is also a preservation concern. Decompilation and recompilation efforts often produce documentation, reverse-engineering knowledge, and tools that can benefit other projects. A rushed port that does not add dependable research or reusable infrastructure may have limited value beyond a momentary demonstration. Critics argue that it can nevertheless receive attention because its output arrives quickly and because AI lowers the apparent barrier to creating a project with a familiar title.

A community response to buggy and untested ports

The backlash extends beyond the Banjo-Tooie dispute. Community member FunkyLion assembled a list intended to distinguish human-made decomps and recomps from AI-assisted alternatives. The goal is straightforward: give players, modders, and prospective contributors a way to identify the origins of a project before downloading it, covering it, or treating it as representative of the broader scene.

FunkyLion argued that generative AI has enabled people with little or no programming experience to create ports that may be insufficiently tested. In that view, the result is a growing collection of buggy projects that can eclipse more careful work from people who have spent substantial time understanding old software and console-specific behavior.

That concern does not mean every tool-assisted project is identical, nor does it prove that any individual port cannot be improved. The broader complaint is about disclosure, standards, and expectations. If a release uses AI-generated code, critics want that fact to be made plain. If a project claims compatibility with established patching and modding practices, they want those claims to be verifiable. And if a port is experimental, unstable, or incomplete, they believe audiences should know before it is elevated as a finished alternative.

For a volunteer-driven ecosystem, trust is especially important. Contributors are often deciding where to spend their spare time. A project with clear documentation, reviewable code, and a transparent development process is easier to support. One associated with unclear authorship or difficult-to-maintain output can pull attention away from the people doing foundational research while creating extra support work for everyone else.

MediEvil developers add their voices

The debate has reached beyond Nintendo 64 projects. Jay Gunn, who is working on a new recompilation project for MediEvil, publicly supported the opposition to AI-driven "slopcomps." Gunn said he did not know more about the controversy than what had been posted publicly, but agreed with the sentiment that generative AI should be kept away from MediEvil now and in the future.

Gunn's comments point to a cultural dimension behind the technical dispute. Older games were made under severe hardware limitations, often with bespoke tools, unusual workarounds, and highly specific engineering decisions. Recompilation work can be a way of studying and preserving that craft. For many developers in the scene, the appeal is not just making an old game run better; it is learning how it was made and building a respectful bridge between that original work and contemporary hardware.

Generative AI does not automatically erase the need for skilled engineering. But the criticism emerging around these projects is that using it as a shortcut can obscure the human labor that makes reliable recompilation possible. The loudest voices in the community are asking for more than a working executable: they want accountable development, transparent methods, maintainable code, and respect for the research that came before.

What happens next for recompilation projects

No single community list or developer statement can settle the question of how AI will be used in fan development. Tools will continue to evolve, and individual creators will set their own boundaries. But the reaction to Banjo-Tooie Recompiled shows that technical communities can still establish norms around attribution, labeling, testing, and project stewardship.

For players, the practical takeaway is to look beyond a project's title or a short gameplay clip. It is worth checking whether a port identifies its developers, explains its methods, documents known issues, and supports the modding workflow it advertises. For contributors, the dispute may reinforce the value of slow, documented progress over quick releases with uncertain foundations.

As unofficial recompilation work expands to more classic games, the divide may become increasingly visible. One path prioritizes rapid output, even if the code base is difficult to assess. The other prioritizes human research, careful implementation, and infrastructure that others can build upon. The current argument makes clear which side many of the scene's established developers believe will keep these projects useful long after the initial excitement fades.