GZDoom developers split from creator amid AI controversy, launch new fork UZDoom

Daniel Sims

Posts: 2,422   +73
Staff
What just happened? GZDoom has long been one of the most popular source ports for playing classic Doom games and their various mods. Its impressive flexibility also supports standalone games that barely resemble id Software's 1993 classic. However, tensions between its developers have led to the impending release of a new version that could introduce significant changes.

Many developers of the popular Doom source port, GZDoom, have recently cut ties with its creator and plan to release a separate fork of the project, which could alter the course of its development. Early source code is now available, and compiled builds are expected to arrive within days.

A GitHub page for the new source port, currently called UZDoom, simply describes it as a modding-friendly OpenGL and Vulkan-based port of Doom. However, early discussions among the project's developers indicate that it will continue the work initially planned for GZDoom version 5.0.

This GZDoom schism has been a long time coming, too. And I think it's for the best, because all the good engine coders are already lined up to work on the newly branded UZDoom. Folks adding better netcode, raytracing support and other features that have been long overdue.

– Dominic Tarason (@dominictarason.com) October 14, 2025 at 7:52 PM

Upcoming features might include improved netcode, ray tracing support, and more. The developers have already confirmed that, unlike GZDoom, UZDoom will deactivate texture filtering by default, a feature many consider unsuited for Doom's original art direction.

The schism between GZDoom's developers occurred after the source port's creator, Christoph "Graf Zahl" Oelckers, allegedly introduced AI-generated code to the project, angering his development partners. They argue that using a large language model likely violates GZDoom's GPLv3 license, since AI-generated code cannot be copyrighted. Although the source port is typically used for free mods, it also forms the basis for some commercial games, such as Selaco and Hedon, which could encounter legal trouble from incorporating unlicensed code. The breakup's impact on their development remains unclear.

Major evolution in the GZDoom community tonight - the regular contributors finally had enough of the project lead, who had been very hard to work with for a long time. He dug in and told them "If you don't like it, go and make your own fork." UZDoom is now live here. github.com/UZDoom/UZDoom

[image or embed]

– AUTOEXEC.BUN (@davidxnewton.bsky.social) October 14, 2025 at 8:56 PM

After GZDoom's developers accused Oelckers of other unprofessional behavior over a long period, he invited them to fork the project, which they promptly did. Some are preparing statements to describe what happened in greater detail, but have begun publicly mourning a source port that recently celebrated its 20th birthday.

Doom mods and source ports proliferated after id Software released the iconic first-person shooter's source code in 1997. One of the most popular, ZDoom, emerged the following year.

Another fork, GZDoom, saw its first public release in 2005. GZDoom is one of the most popular methods for playing Doom on PC due to its support for modern rendering features and the removal of many of the original game's technical limitations.

Permalink to story:

 
I'd argue other source ports have surpassed GZDoom (though it really does depend on how much you want things changed; I've always been a Crispy Doom guy myself), it has been one of the foundational source ports for a VERY long time, and many ports derive from it in some way. Sad to see it go, but yeah, the GPL likely makes it incompatible with AI generated code.
 
That's weird. Been using zDoom since 2003. Drama is nothing new to the Doom community. Can't understand why anyone would need to use AI generated code for a doom port though.
 
Interesting. I'd say GZDoom's development has already been a mess for a long time - at least, since the disastrous changes in the lightmap rendering 2 years ago, a change that no one asked for and no one wanted except the devs, in name of "muh accuracy". It made lighting worse and broke lighting in lots of mods. Another typical case of dev narcissism making a project worse.

I hope this new fork brings back the old lighting as an optional toggle at least.
 
That's weird. Been using zDoom since 2003. Drama is nothing new to the Doom community. Can't understand why anyone would need to use AI generated code for a doom port though.

The guy's just a petty tyrant & an AI truther, simple as. He likes the idea of never having to do anything but have an idea & see it happen, even if the code that makes it happen is unworkable crap.
 
AI‑generated raw outputs are in the public domain (their copyright is null) because they lack human authorship and anyone can freely access the information they provide—this is a fundamental right. In their raw form, such outputs cannot be copyrighted—much like code released under a MIT or Apache 2.0 license that explicitly grants broad reuse rights while still preserving the author’s copyright..

When you integrate them into other programs or codebases, the resulting compilation may be copyrighted. This protection stems from the originality of your selection and arrangement of the AI‑generated material, not from the AI itself. It is generally thinner (more limited in scope) than the protection afforded to works created entirely by a human author, because the underlying elements remain free for anyone to use.

The bar for copyright protection is very low. Even seemingly trivial works can have a copyright holder. For example, a catalogue of names, addresses and telephone numbers is itself copyright‑protected despite the underlying facts being non‑copyrightable; the protection comes from the way those facts are chosen and organised, even though almost no creativity is required.

This illustrates just how low the threshold for copyrightability is—once you apply any minimal creative effort to arrange or combine information, you auto gain copyright and exclusive rights over that compilation. The only restriction is that the work must fall within one of the recognised copyright categories (literary, musical, dramatic, choreographic, pictorial, graphic, sculptural, audiovisual, sound‑recording, architectural, software). The learned numerical values of a neural‑network weight file are not part of any copyrightable category, so they are not copyrightable. Consequently, no one can publish them under a restrictive license or claim copyright over them, because that would be fraudulent (at minimum). This holds true even if the model was trained entirely on a personal dataset; the form of the weights still remains non‑copyrightable.

In short, the AI (which is transparent to the law) belong to all collectively but not to any individual and it provides free ingredients for further creations.
 
Last edited:
The guy's just a petty tyrant & an AI truther, simple as.
Over-react much? Calling the guy a "tyrant" is a serious over-statement. Your statement says much more about you than it does the person you're making the statement about..
 
Back