Skip to main content

On Sale: GamesAssetsToolsTabletopComics
Indie game storeFree gamesFun gamesHorror games
Game developmentAssetsComics
SalesBundles
Jobs
TagsGame Engines

Excellent question! Code, voiceover, etc. also count as game assets and are subject to the same rules. ChatGPT's training data is a black box, and is currently not allowed. I am having troubling finding provenance for Novel AI and DeepL models, so I can't say they fall inside the guidelines either. (If you have more info on them, would love to see.) It looks like Replica Studios has properly licensed and paid for their models, so that would be allowed. (Although if you have time afterwards, I'd encourage you to look for some free VAs for this anyway - I've been told they're fun people to work with.)

Thanks for the clarification! Unfortunately, due to time constraints, I won’t be able to make it this year without these tools. Hopefully I’ll have more availability next year. Thanks again for the quick response! And good luck to all participants!

I'm sorry to miss your work this year, but I hope to see you next time!

Did I get right that drawing AIs are not allowed either?

Correct!


*Unless you own the training data in its entirety.

(3 edits)

"AI-generated assets are only allowed if you own the training data. (e.g. Photoshop's content-aware fill, training an entirely separate instance of AI on exclusively your work, using ML to remix art with permission from the artist, etc.) Systems in which the training data is not owned by the dev (ChatGPT, Midjourney, Stable Diffusion, and similar) are not allowed at this time" [From NanoReno page]


An asset - property is anything within the game itself as clarified already by sake. This includes, both physical assets (VA/Art/Code) as well as IP assets (Ideas, style etc).


This is a bit of a quagmire, and I think really the rules should be updated to just ban use of prompt based AI outright.  As the exception is a bit of a grey area. Also when you license the use of an AI, you are potentially also licensing the training data, which is a legal argument I really don't want to talk in detail about as its a waste of time pivoted off of legalese and semantics.

This is definitely a rule that will be reviewed more thoroughly as the tech and legal areas around it evolve. I like your suggestion, and I think it will likely be the one the community lands on.

(1 edit)

I think the current rule is pretty fair. I don’t see why there would be a reason to ban things generated with your own data or royalty free data using a diffusion model. the current rule touches on the main issue which is consent to use that IP. Currently the hosted solutions Midjourney, Stable Diffusion, NovelAI its not publicly known if they have artist consent for their models and ChatGPT definitely doesn’t have peoples consent because its trained on web scrapping. Banning prompt based AI excludes perfectly valid uses of the technology.

(1 edit)

Hi Stag

The problem with AI generated content is that it is low effort content and does not showcase your ability nor improve your ability through process learning, that is because AI generated content is not your work, even if it is a model trained on your work. So, it actually misses the point of what the jam is all about, which is improving and showcasing. It is the same reason that AI generated content (prompt based) does not garner copyright attribution regardless of the source material used in its data sets. At best you might claim it created a derivative work based on the prompt itself, but the argument is very poor. 

Non-prompt based AI is mostly QOL stuff, for example smoothing algo, or animation automation (you specify certain frames and in fills in the gaps), and have been long in use for yearas prior to this every becoming a thing.

You don't need artist consent to train an AI off of their work as the AI does not do anything to breach copyright. However, if the AI produces something far too similar to their work that in and of itself could constitute a breach, but this doesn't preclude all AI generated assets from said data training.  So, there is nothing invalid from a creator using prompt based AI.

However, prompt based AI is something I believe does not follow the spirit of the jam. Hence my wish/suggestion that all prompt based AI was banned.

Hi Tremarl ☺

the copyright ruling is very new so I don’t think its set in stone but I think we agree on the legal and consent based elements. I mentioned consent because its a common point that makes people uncomfortable and by making people train there own AI makes the task no longer low effort.

I agree that AI generation makes it easier to create aimless art with premade models but its not easy to create artwork with purpose like you need for these novels. I agree AI Art doesn’t teach you how to be a digital painter; using AI Art generation is closer to the role of an art director. The jam allows purchased & premade public assets so improving and showcasing your skills as an art director or digital artist or vroid character designer or metahuman designer or any other art abstraction are all part of the spirit of the jam in my opinion but lets wait and see. I hope our differing points of view can help the hosts of the jam in their choice of rules and I hope you have a great jam 🫂!