The explosion of AI into the homebrew scene has had mixed results, to say the least.
On the one hand, the 'democratisation of art' argument says that developing games is no longer the sole domain of the technically accomplished coder, or those with the patience and persistence to learn one of the many game development systems (Unity, Godot, RPG Maker, Adventure Game Studio, Ren'Py, Scorpion Engine... to name a handful off the top of my head). Some of these are even made available for free, to encourage development of new homebrew games. But now, one can 'vibe code' an entire game, create assets 'for free' within another AI framework, obtain AI-generated music and sound effects, etc. all without experience or expertise in any field other than describing an idea, and in a fraction of the time it would take a dedicated team of produce.
On the other hand... this very freedom and accessibility has led to an explosion of (I'm going to try very hard to avoid the term 'AI Slop') hastily packaged, vibe-coded wares thrown together for fun and profit by people who have clearly been angling for an easy route to make money or gain some online attention with some bad ideas. A prime example of this happened recently, with the Steam Next Fest finding itself inundated with software created with AI, much of which was lazy AI 'clones' of projects by genuine developers, some of which had barely even been promoted. Perhaps the most egregious example was the 'Needle in a Haystack' game, which suddenly became an entire genre with dozens of variants on Steam. This led some of those developers who'd hoped to showcase their projects as part of Next Fest deciding to pull them completely, and more strictly gatekeep access to their demos. Given that online promotion is the only way most of these projects find an audience, they are essentially running the risk of scuppering their games in order to protect them... And that seems to be the only reliable method right now, because there's one game that had been promoted only with static screenshots, yet still got ripped off by a Roblox user before it had launched.
On the other, other hand, AI has facilitated ports of games from and to various formats, both retro and new, and surely that has to be good thing? The one consistent problem indie developers have is a lack of time to work on their pet projects, so using AI tools has to be a wise use of newly-available resources to make quicker progress and, ultimately release that project far sooner than they otherwise might. As long as they're not relying on it completely, and as long as they understand the code produced, at the end of the day, they can still feel a sense of accomplishment when they finally release the project, right? And they'll be in a better position to start their next project, whatever that may be, right? And if there's a particular conversion they've been aching to work on, they're now in a position to make it happen far more quickly, right?
Well, here's where things get awkward:
Early in September, a member of the SAM Coupé Users' Group on Facebook posted a collection of games he'd ported to the SAM using ChatGPT. This was swiftly followed by two more posts, each of a single game converted by the same method. The next day, another batch appeared. Two days later, another batch. Before the end of September, a total of 20 posts had been made, comprising AI-facilitated conversions from the ZX Spectrum, Amstrad CPC, Commodore 64 and MSX. Most of them were ports of games from the big names of the 1980s and 1990s, but there were also a handful of homebrew games... One of which was Westen House.
That's the same Westen House that the Lower Caverns coder and I have been working on, with the blessings of the original MSX developer, and first announced at RetCon back in 2024.
This conversion is a bare-bones port of the MSX game, using its original graphics, but in the SAM's MODE 4 to eliminate what little colour clash there was. The graphics haven't been altered, the music is untouched other than to make it suitable for the SAM's SAA1099 and, to be frank, the game runs far slower than it does on the MSX. All of this done with no apparent regard for the work officially in progress for the SAM conversion making full use of MODE 4, with graphics lovingly adapted by yours truly.
Mercifully, perhaps, this occurred during a period when the coder was largely off social media for health reasons, and a couple of other members of the Users' Group - myself included - politely suggested that the admins might want to remove this game, if not a whole bunch of other unauthorised ports of recent MSX homebrew games. As I write, they have declined to do so.
However, the coder returned to social earlier this month and was understandably angry, not only about Westen House receiving an unauthorised and lazy AI port - which almost made him throw away his own work - but because there were other games in this library that, while possibly not in active progress, were on the wish lists of a number of other time-starved SAM coders who might now give up on their plans.
Worse still, the other homebrew games - particularly those on the MSX - could have been developed legitimately, the same way as Westen House: by contacting the original developer and asking their permission. Instead, they have effectively been pirated for the sole purpose of turning around a quick port to show how useful AI can be... And this, for a machine that has a huge back catalogue of Konami games ripe for conversion to the SAM, because Konami don't seem to give a damn about much of their back catalogue. I've even posted mocked up SCREEN$ of Metal Gear on the SAM, suggesting that it would be plausible. There could have been a port of the Gradius and/or Parodius games but, no, the AI mill decided to grind through indie games - passion projects, often from solo developers - from the last few years.
And as if that kind of piracy wasn't enough, the same Users' Group member decided to port a Lucasfilm game, and the SCUMM engine - itself essentially a homebrew project aiming to keep retro adventure games playable on modern systems... but (technically) first requiring that one actually own a copy of the game they wish to play, many of which are long out of circulation. This port packages the SCUMM engine along with the C64 version of Maniac Mansion.
Depending on how litigious some of these people are, this could have significant repercussions for the SAM, despite its small userbase and niche position in retro circles. Knowing that someone has used AI to bodge a port of one of your games to the SAM, how would you feel if someone contacted you about converting another one? Knowing that a 'no' would likely result in an unauthorised port, I'd still be inclined to say no... but then, two of the projects I'm covering here (Westen House and Reckless Rufus) have come about through dialogue with the original developer, because that just seems to me to be the best way to work. If you hit a roadblock, you can ask the original developer if they had the same experience and how they got past it.
Supposedly, there has been something of a misunderstanding of intent over this. At the very least, the bare bones MODE 4 adaptation of Jet Pac is supposedly a work in progress that will eventually feature hand-crafted sprites by one of the other artists on the SAM scene. Additionally, at least one coder has conceded that they would likely have used the Spectrum's visuals in MODE 4 as the basis for a SAM conversion, with a view to replacing them later on, were they to attempt a conversion by more traditional means. In that respect, isn't it fair to say that AI has simply shortened the development time to however long it takes the artist to produce the unique SAM graphics and for the AI-assisted developer to patch them into the game? That's all well and good, but the AI-assisted developer in question has neither mentioned this intention nor engaged in the debates following his postings. Granted, English is not his native language, and it was his artist who first revealed the claimed intent behind the port of Jet Pac. However, this being an Ultimate game, there's still a risk - however small - of repercussions for a port that is both unauthorised and AI-assisted.
And the bottom line on all of this is that the SAM scene is so starved of good games - and pretty much has been since the very start - that I can't really blame anyone for taking advantage of AI tools to quickly produce ports from other machines... I just wish they'd focus on legacy titles rather than games created less than five years ago by indie devs.
I have to admit that, as a fledgeling developer myself - aside from being a pixel artist on the SAM, I'm trying to create a series of Visual Novels using Ren'Py, with tentative plans for a side-story in RPG Maker. I've also downloaded DizzyAGE, the Scorpion Engine and Adventure Game Studio, with grand plans for each once I have the time and motivation to learn how to use them.
AGS might - just might - eventually lead to a certain point-and-click adventure I've had rolling around in my head for over thirty years... but that still leaves an awful lot of graphic artwork to produce.
Scorpion Engine seems like a good development environment for a fudged version of my Empire Strikes Back concept, though it's unlikely I'll be able to make it all work the way it's intended all by myself... at least, not within the current decade..
The idea behind DizzyAGE is to create a RetCon-themed Dizzy game for a future RetCon and, while that sounds like a more reasonable project on the surface, my brief dabblings suggest that even DizzyAGE is quite an involved system.
The AI-assisted route could, conceivably, help move any of these along at a faster pace. It might even be possible to vibe-code my Empire Strikes Back concept onto the Spectrum. Apparently AI has a decent grasp of Z80 code, presenting solutions that have impressed long-time Z80 coders... And the fact that I was met with a very chilly response from the Facebook Spectrum community tends to suggest that's about the only way I could make it happen for the Spectrum.
But, much as I might want to give it a try, there are some things I just don't want to use AI for, however much easier it might make the process.
Your mileage may vary.
There's a wide spectrum of opinions about the use of AI, from the knee-jerk "AI is evil! It should NEVER be used for x..." to the more nuanced "It's a tool... and it's literally a part of human nature to use tools to make things easier for ourselves."
For me, though, learning to make the most of those tools and refining my manual processes is part of the fun, and I see improvements in my work since I've been using Aseprite as my main tool.
AI is throwing other people's work into a blender, then using that to present what it thinks I want. I've had no involvement, but other people's work has been scraped and regurgitated as new. If that's what I wanted, I could use reference material and still do the drawing myself - and that's another trick I've been using more frequently since working in Aseprite and after how well The Lower Caverns turned out by using the GameBoy Advance version of Manic Miner as reference. To me, prompting AI to create graphics isn't 'using a tool', it's abdication.
With my somewhat strained relationship with animation, the tools out there for getting AI to create a sprite sheet for various common actions are somewhat tempting. My 'method', such as it is, is slow and labour-intensive, but I'd still rather work from references than let AI tackle it all.
The coder on the official SAM version of Westen House is adamant that he's never going to use AI in his development, and I've put together a 'No AI' SCREEN$ for him.
On a related note, in my current job, I've dropped in on the Adobe User Community quite regularly and, while there are dozens of users furious with Adobe for going all-in on AI features when the base applications are still buggy and unstable, there are some who are experimenting with those features. One such user described a workflow where he'd sketch an idea, run it through AI to flesh it out, to provide detail and options, which he'd then redraw in his own style. While that seemed satisfactory and beneficial to him, the AI iteration stage seemed superfluous to me.
And, tangential to that, there's there emergence of the Craft suite of apps - essentially open-source versions of a set of Creative Cloud apps, created with AI assistance. Since a good portion of the posts in the User Community are complaining about the excessive and cancerous AI features in Photoshop, InDesign, etc., all of which were enabled by default and spending users' credits without their knowledge, and the bugs that have persisted, unaddressed, through dozens of software updates, the idea of apps that look and feel familiar enough to Adobe's, without being stuck behind a draconian subscription model, without the AI features and with bugs actually being dealt with as they arise, is mightily tempting. My girlfriend has questioned the wisdom of downloading early versions of AI-spawned rip-offs of professional software on the grounds that one can't be sure there isn't something dubious going on behind the scenes... but development of the Craft suite continues (pending legal action by Adobe that has the potential to backfire on Adobe themselves and the wider AI ecosystem), and as its reliability is demonstrated, Adobe stands to lose a significant proportion of its userbase who are currently shackled by their Creative Cloud subscriptions.
The AI question is far too big for one offhand post, particularly as its proponents and developers attract more media attention (and lawsuits) every week. This post originally started out as something related to a completely different project, before the AI-assisted port of Westen House appeared, so I'll probably continue my thoughts on the subject when I have more to say about that one.
Comments
Post a Comment