AI art and AI in general is a hot topic nowadays. Has anyone tried to upscale game sprites and other images for MUGEN? I would recommend checking a free open source tool chaiNNer. There's also a large database of various freely downloadable AI models for upscaling and people even can also train AI models themselves. A lot of paid services popped up too for an easy cash grab, but I'm not sure if they can really offer something that isn't already available for free.
Here's a screenshot of the free chaiNNer app:
The AI doesn't just produce larger blurred pictures. It uses "creativity" and "imagination" to construct the missing details not present in the original low-resolution picture (on the left side of the screenshot), such as proper eyes or hair. But it's still not very good at drawing human hands. It's possible to try different models, pick the best picture and edit it if necessary. What do you think?
I'm not sure folks would appreciate that, as a person in the art community with twitter, if you were in my position you wouldn't go 5 minutes without seeing someone calling out a company for using AI and telling people to boycott them. So, if someone was to upscale Mugen sprites, I doubt it'd go well. Would def be interesting to see how it'd work, since iirc it doesn't take too kindly to high-res sprites, so maybe you'd have to apply it in game rendering? I'm not sure, I don't have much programming experience.
Interesting, I have not had the opportunity to use AI in that aspect (only academically) but I had and have faith that it can give very competitive results. In the example you share, the sprites look like a "High Res" style, which can give new life to already known characters.
I think that a tool of this type has a future in this MUGEN environment, but the limitation would possibly be that it is only a character already created previously but seen "more current in the drawing" but it would have more future to develop original characters, perhaps more limitations would be the cost of using a more "better" version.
As for the artists/illustrators, although it may be a competition, perhaps this is a window of opportunity for everyone, I say whoever has the idea can now capture it as is without having to go looking for the artist who wants to carry it out " because some set their conditions" or high costs, be careful not to say that I do not value people's work but sometimes the costs seem disproportionate to what is required, but I suppose that by searching you can find it, but the fact is that First-hand there is the tool to create without complications in terms of sprites.
I don't demonize the use of AI, since it saves a lot of time, which is the most valuable resource we have, and it has proven to yield interesting results. For example, with animations, we could create complete characters or remaster things that previously had very low resolution. Can you imagine a game with characters like this, for example?
[media]https://grok.com/imagine/post/455076e3- ... atform=web[/media]
Atreyu888 wrote: ↑Fri Feb 27, 2026 1:55 am
I don't demonize the use of AI, since it saves a lot of time, which is the most valuable resource we have, and it has proven to yield interesting results. For example, with animations, we could create complete characters or remaster things that previously had very low resolution. Can you imagine a game with characters like this, for example?
[media]https://grok.com/imagine/post/455076e3- ... atform=web[/media]
Really depends on, because Ai content falls a lot on the grey and black scale rather then white and blue.
A good example that no one ask is MicroSlop, enforcing Ai to be used while monitoring everything you do to train their Ai product.
Another one would be Movies, and CallofSlop, Not Call of Duty, but Call of Slop due to adding Ai images inside the game.
I remember the early days of AI upscalling before ChatGPT made AI/hating AI popular, that's right I'm talking about ESRGAN. The problem I had with it then and have with it now, is that really it usually just results in a blurry mess.
It's not just from amateur use either, you see this same problem with (for example) Square Enix "HD" re-releases of the old Final Fantasy games, the problem was so bad that people cheered the switch to "Pixel Remasters". You can see the same problems in the demo picture of this thread. Look at Karin's face on her sprite, her face is always a blurry mess. There's a whole genre of characters like this on Mugen archive {just search UPSCALE, you'll get a whole lot of [Character] HD (UPSCALE + SHADED)} now I will be fair and say the portrait generation was a lot better than the sprite, the middle portrait in particular is practically perfect.
Now AI upscaling can be good, if it's heavily baby sat. Awhile back I downloaded a fanmade remaster of the 1999 Hunter x Hunter series, and that looks perfect because the team working on it heavily revised it (I think I got version 3 of their upload). For the most part, you can't just "set it and forget it" with AI upscaling if you want to make anything other than a 'Blurmaster'
