What do you expect to hear here?
Obviously you can try make game and see it for yourself, if you make it crash.
It won’t change a fact, lots of people wants avoid AI slop here.
1 Like
Ok to be more transparent in Open Bor, if you sprite was not an 8 bit indexed-256 color palette the game froze, crashed whatever the engine choked. AI images left vague logs. I realize more modern engines have better capabilities, than others but it’s primary engineer, who is not my favorite type, claimed that using 2D Ai sprite animation efficiently wasn’t possible most open source engines including Open Bor, Unity, and Godot. I used to run 22 frame animations on it but he claims I would encounter problems using Ai sprites on any modern engine .
He might’ve been referring to the ethical problems and the fact that most people will not be willing to help you if you use AI for art.
2 Likes
The question you are asking remains vague.
The answer* (imo) remains “just try it out”; if it works it works.
This remains the answer because bandying about here serves no purpose. What could anyone tell you about images that hasn’t already been said and that we don’t have access to?
An AI generated .png file is going to be the same format as you draw one manually so what difference does it make?
I had a look at Open Bor and can see nothing that would prevent you from using AI generated assets.
I can only conclude that there must be a lot of info we are not seeing.
*Unless the real question is “how do I implement a pipeline direct from AI to Godot so that my images just work?”. Obviously that is a much more complex question.
PS: The Godot engine is already 12 years old; far from new as far as tech. goes. OpenBor engine latest release looks like May 2025.
Here is what Gemini has to say about it (Warning AI content in the quote, avert your eyes mortals!):
Yes, you can absolutely use AI-generated images to create a 2.5D game in the OpenBOR engine. However, because the engine relies on traditional 2D sprite sheets rather than live 3D rendering, you will need a specific pipeline to convert your AI artwork into OpenBOR-compatible assets.
I read that to mean that you must turn your AI images into a sprite sheets formatted specifically for the engine.
With Godot you would want to do something very similar just to make things a lot easier for yourself.
1 Like
If by ethical problems you mean fan art 75% of his games/mods are Capcom reworks. Their most popular games are a He-man and a GI Joe game . In fact his most active member constantly remakes old Marvel games. I never asked for his help or had to show him what I was working on. This was a simple debate about can AI sprites of any character be used efficiently believe me that’s not it. He was always too concerned with paragraph structuring even AI is aware he pushes more advanced developers away.
Define “Ai sprite”. What does that mean?
yeah most Open Bor modders -some game creators, use sprite sheets because they want to make a game or mod but do not necessarily want to worry about creating art too. Speaking as a developer most professional software companies have an art department, an engineering department, sound, executive and conceptual designers. Plus legal departments. For every game a developer releases in 1 year 20 get scrapped. Because of budget and time restraints and even licensing issues
Development takes time and teamwork
Back in the 90’s for instance. Capcom Street Fighter 2, for every frame artists would draw 2000 or more just to get 100 animation frames if that.
For 2D or 2.5D basic late 90;s arcade games which Used 2D sprites. Open Bor is one of those engines Open Bor characters animation looked like this. that’s 80% animation smart 20% stupid code. That’s just the way it was designed
anim idle
loop 1
delay 5
offset 35 34
hitflash shock
hitfx data/sounds/jolted.wav
shock 17 19 37 31 10 1
frame data/chars/knbot/shball1.png
frame data/chars/knbot/shball2.png
frame data/chars/knbot/shball3.png
frame data/chars/knbot/shball1.png
frame data/chars/knbot/shball5.png
frame data/chars/knbot/shball4.png
frame data/chars/knbot/shball1.png
frame data/chars/knbot/shball2.png
frame data/chars/knbot/shball6.png
frame data/chars/knbot/shball1.png
frame data/chars/knbot/shball5.png
frame data/chars/knbot/shball7.png
Now some purists such as the earlier poster look at AI as a cop-out and certainly AI generated art can’t be copyrighted but I have worked in the industry my whole life and I can tell you there’s a 100,000 people who if had AI 30-40 years ago would have loved to have made their own jobs easier or for private fan projects not interested for marketing but just something they wanted to do behind their own doors because they bit or held their tongue because they saw a mistake happening but had no say. I am an engineer not an artist and for me AI is a dream at my age you do what you know and my job was always giving NPC’s brains.
I’m not looking to import for bor, Id like to create something I coded without needing an artist. In conceptual design you can often describe to AI what you want to see faster rhan an artist can draw it. I don’t want to see the artist replaced I have known dozens but for once in my life having art freedom for my own private project isn’t going to damage anyone.
I have pretty much retired but I often said with my programming experience I’d love to create something for myself to say without pressure or budget constraints AI combined with manual adjustments has opened that work up. And I believe Godot could be the answer.
In Godot you can load animation frames from a file sequence as well as from sheets/atlases. Using sheets may be better for performance. Making sheets from file sequences is trivial. You can use one of many utilities available around the web or write your own Python script that does it.
You can’t use indexed color images (like gifs or indexed pngs) directly in any modern engine, including Godot. Today’s graphics hardware doesn’t natively support indexed color modes any more. However, you can import indexed pngs into Godot and the engine will re-rasterize them into standard rgba color mode on import.
There’s plenty of “ai” generators that claim to be able to create production grade animations/sheets from a single character concept image or verbal prompts. However they may be performing sub-optimally to a large extent. We recently had a poster here trying that out and abandoning some supposedly top-tier tool because they couldn’t get a good enough animations out of it.
So your problems will likely boil down to aesthetics and other artistic and functional aspects of animation, rather than technicalities with handling images.
1 Like
You should probably look up the concept of straight and premultiplied alpha.
The things you describe, that are related to alpha premultiplication, are not specific to “ai” generated images. Same can happen when doing classic 3d renders, and indeed when drawing sprites by hand. In any case, engines can handle this, but they’ll prefer premultiplied alpha, for the reasons described here. Be warned it’s a bit of a demanding read:
https://www.realtimerendering.com/blog/gpus-prefer-premultiplication/
All of that can be “fixed” with simple image processing scripts or with available tools. That also goes for removing unwanted metadata chunks from image files.
There’s no problem whatsoever with any of this. Again, your main problem will be getting those “ai” sprite generators to keep up the consistency.
I’d go deeper into finer points of all this but I don’t like to respond to LLM generated text. When you post “ai” text without disclosing it, watch out for that signature rhetorical form that at this point even my auntie can recognize: “It’s not that X (em dash!) it’s that Y”. It annoys everyone.
5 Likes
(post deleted by author)
Only the slanted and *text was pasted the rest I wrote. Its on my website Under Q%A although I did use AI for where there was a*. I do that to shorten. Consistency hasn’t really been much of a problem after studying output generation save for a few small sprite surgeries. I have 5 characters fully animated. I did encounter problems. Its typically addressed through understanding what caused it. You jjust need a good anchor. I always begin with the same reference , being that it often generates the same frame so all I have had to do is dump excessive frames and in some cases return to Idle 1 frame. I have had to do some patching. A simple example was sheathing the sword and reversing for unsheathed although when returning the arm I had to create a few frames. Direction, appearance, angle, had to be maintained. Locking in a workflow is possible but It takes trial and error and referencing. In fact I have only had to intervene about 25% of the time. You always have the option to recreate. You only have the option to recreate.
I wouldn’t go that far. If anything it’s like a collaboration of thousands images . There are certain characters it won’t generate for creative purposes if it matches 60 percent.
Answer
I reposted this because I wrote every bit below verbatim. I included the Ai interpretation because often it might include some technical elements I might not. I was not aware that all AI text was banned.
Usually in PNG format generated by AI is not like a sprite created by a human artist. AI sprites use RGBA and human artists use RGB sprites. "The difference is AI sprites often put gray translucent pixels around characters which engine developers may or may not have to remove. RGBA sprites often include chunks of various Metadata which each engine resounds to differently. Modern engines usually have no problem. Older engines do not recognize what metadata is causing them to choke and maybe crash the game.
Because it was created through a different method simply resaving it as a png or converting it doesn’t always change that it was generated as an RGBA sprite. You could put it Gimp and its still different than RGB. Here’s a way to show you better. Take any magic tool around the character in a background it will often -if you zoom in leave little gray translucent pixels around the character and possibly a chunk of background. I have spent hours zoomed removing these often you may not see them zoomed out. Even updated open bor and older engines may experience issues because of these. There is one AI background remover that typically gets it all.
You cannot be serious.
4 Likes
Yeah it searches millions of databases and draws an artificial inspiration combined with influence from the prompt if you request it generate a black dragon, with specific features it doesn’t go to a website and highjack someone’s painting. you determine details. i have had it draw specific charters, sure I’ve also request it turns my drawing into pixel art. i suppose you could say its a giant array of templates but then at this how many different ways is there to draw a dragon.
It’s really unfortunate that most big companies has successfully convinced people that AI generated art is just “magic” in that regard.
“No it absolutely does not steal from other people.”
Yes. Yes it does. At the time of generation, of course it won’t “go to a website for someone’s painting”, because it already did all of that during training. Most companies simply took EVERYTHING they could get their hands on from EVERY single website out there to train their models on. Without permissions. That is just blatantly stealing.
And this is not just with art, even Meta got into trouble for illegally torrenting a massive amounts of books to train their AI on.
5 Likes
Thanks for the link I solved this old or new no engine that uses frame by frame sprite animation can upload RGBA because they always anticipate RGB which can’t be read. Thanks though .
I don’t know what you mean by old books. But I have only used Meta for random general information, never prompted books before . I have prompted creatures I have never seen before and detailed 50 random aspects of it.
Where did you get old books from? And how is that you haven’t prompted books relevant to this? The fact is still that these tools plagiarize and do not create
(Also completely unrelated, what is “Go dot”?)
3 Likes
A whole bunch of answers that directly contradict you. You seem intent, with confirmation bias, to be proven right in that misbegotten view of yours.
The provenance of the image doesn’t change how a game or graphic renderer uses it.
And you can use RGBA and RGB images, in various color profiles, SDR and HDR, sRGB or not.
3 Likes