Clair Obscur is not indie by any definition of the term. I don’t even know why it was considered at all.
Sandfall *interactive is independent from its publisher Kepler. Many of the other games Kepler produces are typically considered indie - why not Expedition 33? BG3 is “Indie” but this definition
While Hades, Hollow Knight, and Celeste being both owned and published by the same company are not indie.
So… idk what definition everyone is using. Seems to be whatever suits their agenda at the time of award.
While Hades, Hollow Knight, and Celeste being both owned and published by the same company are not indie.
if your definiton of inide exclude Hades, Hollow Knight and Celeste because they are independent i have to say that it is a very bad definiton of what an indie game is.
personally, if a game has enough budget to hire Charlie Cox or Andy Serkins, it probably should not be in an indie award ceremony
Yes okay but how do you define it?
Because that is all that “Indiependent” means.
Remember Hades and Hades 2 had a bigger budget than E33
- Hades production cost was over $15 Million
- Hades 2 production cost was over $20 Million
- E33 was less than $10 Million.
Hollow Knight was developed by 2 people from 58,000 Kickstarter.
I loved the game.
I understand the use that was made did not in the least affect the final product.
I don’t think they should have a disclaimer on Steam.
I think they screwed up big time if the indie game awards rules could have been interpreted as requiring no use of AI at any stage in production.
Also, I dont really understand the point of saying it afterwards and I fear that may in itself mean that they are promoting the use of AI in game dev.
What I think is very good is that people are (over?)reacting like this: I would like to have devs perceiving the use of AI as fucking poison.
Clair Obscur: Expedition 33 launched with what some suspected to be AI-generated textures that, as it clarified to El País, were then replaced with custom assets in a swift patch five days after release.
Fuck using Gen AI to replace human-made art, and fair enough for pulling the award, but I do think it’s worth making it clear exactly how much of the art is/was AI. And the answer is, very little at launch and none currently.
Also based on development timelines, these couple of textures were likely from the DallE-Mini era, long before anyone was even talking about ethical concerns with the tech and before fuckass Altman was a known name.
There’s another article going around where they say they use a “little bit” of AI but they’re very obviously referring to Unreal tools and regular old machine learning.
This whole thing is dramatically overblown. Absolutely fuck genAI and every data center should be bulldozed over. But this is a huge case over literally nothing.
E33 should be disqualified for all these awards for not being an indie game (the devs themselves have called it “AA” if I recall, and they had a sizeable budget to burn). Should genAI be tolerated in any part of the development process? Absolutely not. Should a game and studio be completely brought down because of 2022-ish placeholders from before the ethical issues of the tech were common knowledge, which were also quickly patched out in less than a week? Absolutely not.
AI wasn’t used to “replace human-made art”, though.
To me it sounds like the team needed generic textures in big batches, and instead of spending precious designer time on hand crafting them, AI was utilised to allow the designers to focus on actual art they enjoy. I’m a software engineer, not a designer, but if I were given the option to write 8000 classes that are almost the same, or write 5 classes that will take the same effort as the 8000, but actually require using my creative skills… I’d choose the latter, and offload the 8000 boilerplates to AI.
The fact that it was replaced with human made art so quickly suggests that the AI generated ones were meant to be placeholders only anyway.
So instead of buying the textures they didnt want to create, they paid for AI to generate derived versions from stolen art??
Whats the point? Just give the artists the money directly.
But these aren’t textures that are intended for use. They’re filler for development purposes.
It’s like putting a gray box in and fixing it later or putting a TODO in code.
Yes, so they can buy some cheap marketplace textures instead of paying to use stolen content. What are we not understanding here?
But, why for placeholders that aren’t meant for anything? No one will ever see these (intentionally).
I’m very anti AI in games for art but this feels like an arbitrary line in the sand for “support artists”. This feels like the stupid busywork that AI can alleviate. If there’s a large market for placeholder stuff maybe but most of the stuff in stores you’re paying to ship in production normally.
Why purchase stock things that may not be the dimensions you need or need tweaking to make work when you can just have something craft a placeholder?
Why go through the effort of AI then for them? It’s the same amount of work, well less even, you’ll get a full pack of textures at different resolutions with UV maps and all, versus AI where you then have to check it and potentially do more.
It says it all, when purchasing a texture is now “stupid busywork”. You literally type in what you want, get hundreds of results, buy one, put in the game. They used UE5, Epic Games have Fab and it’s integrated into the engine, they even give away lots of free assets there (granted that is probably going to fill with AI slop soon). There’s loads of other options out there too.
This argument of “its placeholder textures” is null.
How is the argument null?
The major rub of Gen AI art is profiting off of others work. If you’re never displaying it and don’t ever want it shown then who cares?
I don’t know the red tape around them purchasing packs. They also may not have wanted to do that if they wanted to check for a specific aesthetic. I don’t know and quite frankly I don’t think this is the hill to die on for Gen AI usage.
I’d rather fight the slop or production use of it vs stuff that will never see the light of day.
No. Instead of making designers work on Brick Texture #7298, they allowed them to work on actually interesting design bits while the necessary textures were placeholdered by AI.
Also, stolen art… The same argument comes up here as with piracy. If I take something you created, BUT you’re not deprived of said thing, then it’s not theft. It is a breach of licence but not theft.
I do agree that some genAI models have very questionable copyrighting issues due to source dataset usage, but, just by creating a model you haven’t deprived anyone of ownership of their property. You haven’t actually done any financial damage to them.
So please stop overblowing the issue and instead begin by pushing for support of artists’ rights to decide if their art can be used by third parties for the purpose of AI training, which is the core issue here. And even go and push for artists’ rights to reserve their art’s training data usage to themselves, thus allowing artists to create their own specialised models with their own style that they can use to offer cheaper art, or even license the use of the model out for money, thereby allowing artists to directly benefit from AI instead of being fervently against it.
You’re also forgetting that most companies like Sandfall Interactive, that work on a budget, have their own designers so they don’t just shop around for artists, even without AI. But without AI it would’ve meant that those hundreds of brick etc. textured would’ve gotten a placeholder that was unsightly. See e.g. Valve’s Source Engine pink-black checkerboard placeholder. Would you have preferred that?
Did you even read my short comment? They can buy a brick texture from one of many marketplaces. Giving an artist money directly. Instead of giving money to use stolen assets. So that argument doesnt hold up.
Did you read MY comment at all?
When you have in-house designers you won’t go shopping around for textures, especially not placeholder ones.
Why are you being obtuse? If the in-house designers didn’t want to make some placeholder textures, they could have used a marketplace instead of AI. Are you just going to keep going in circles?
Why are YOU being obtuse here?
No company with an in-house design team will start trawling marketplaces to spend money on PLACEHOLDERS.
And it’s not like the designers “didn’t want” to make the textures, you donut - it’s that resources need to be allocated, and making minor textures falls on very tail end of the priority list.
At which point they probably had one designer generate the needed placeholders using AI, to ensure they’re good enough for placeholders, and called it a day.
I’ll ask you one better - why are you trying to force companies to go out of their way to spend money? When digital design tools hit the market, would you have been standing in line telling companies to instead hire out actual manual art instead of working with digital tools, if they didn’t have the required in-house resources?
That’s exactly the takeaway I got from it as well.
It seems most likely that those were placeholders that were supposed to be replaced before release but were missed. Once they realized that some were missing, they got them replaced and pushed the update.
GenAI being used for placeholder stuff is arguably the perfect use case, especially for small studios without massive art teams.
Can someone help me to understand the difference between Generative AI and procedural generation (which isn’t something that’s relevant for Expedition 33, but I’m talking about in general).
Like, I tend to use the term “machine learning” for the legit stuff that has existed for years in various forms, and “AI” for the hype propelled slop machines. Most of the time, the distinction between these two terms is pretty clean, but this area seems to be a bit blurry.
I might be wrong, because I’ve only worked with machine learning in a biochemistry context, but it seems likely that modern procedural generation in games is probably going to use some amount of machine learning? In which case, would a developer need to declare usage of that? That feels to me like it’s not what the spirit of the rule is calling for, but I’m not sure
I don’t know of any games that use machine learning for procedural generation and would be slightly surprised if there are any. But there is a little bit of a distinction there because that is required at runtime, so it’s not something an artist could possibly be involved in.
I’m not so much talking about machine learning being implemented in the final game, but rather used in the development process.
For example, if I were to attempt a naive implementation of procedurally generated terrains, I imagine I’d use noise functions to create variety (which I wouldn’t consider to be machine learning). However, I would expect that this would end up producing predictable results, so to avoid that, I could try chucking in a bunch of real world terrain data, and that starts getting into machine learning.
A different, less specific example I can imagine a workflow for is reinforcement learning. Like if the developer writes code that effectively says "give me terrain that is [a variety of different parameters], then when the system produces that for them, they go “hmm, not quite. Needs more [thing]”. This iterative process could, of course, be done without any machine learning, if the dev was tuning the parameters themselves at each stage, but it seems plausible to me that it could use machine learning (which would involve tuning model hyperparameters rather than parameters).
You make a good point about procedural generation at runtime, and I agree that this seems unlikely to be viable. However, I’d be surprised if it wasn’t used in the development process though in at least some cases. I’ll give a couple of hypothetical examples using real games, though I emphasise that I do not have grounds to believe that either of these games used machine learning during development, and that this is just a hypothetical pondering.
For instance, in Valheim, maps are procedurally generated. In the meadows biome, you can find raspberry bushes. Another feature of the meadows biome is that it occasionally has large clearings that are devoid of trees, and around the edges of these clearings, there is usually a higher rate of raspberry bushes. When I played, I wondered why this was the case — was it a deliberate design decision, or just an artifact of how the procedural generation works? Through machine learning, it could in theory, be both of these things — the devs could tune the hyperparameters a particular way, and then notice that the output results in raspberry bushes being more likely to occur in clusters on the edge of clearings, which they like. This kind of process would require any machine learning to be running at runtime
Another example game is Deep Rock Galactic. I really like the level generation it uses. The biomes are diverse and interesting, and despite having hundreds of hours in the game, there are very few instances that I can remember seeing the level generation being broken in some way — the vast majority of environments appear plausible and natural, which is impressive given the large number of game objects and terrain. The level generation code that runs each time a new map is generated has a heckton of different parameters and constraints that enable these varied and non-broken levels, and there’s certainly no machine learning being used at runtime here, but I can plausibly imagine machine learning being useful in the development process, for figuring out which parameters and constraints were the most important ones (especially because too many will cause excessive load times for players, so reducing that down would be useful).
Machine learning certainly wouldn’t be necessary in either of these examples, but it could be something that could make certain parts of development easier.
generative ai is a subset of procedural generation algorithms. specifically it’s a procedural algorithm with a massive amount of weight parameters, on the order of hundreds of billions. you get the weights by training. for image generation (which i’m assuming is what was in use here), the term to look up is “latent diffusion”. basically you take all your training images and blur them step by step, then set your weights to mimic the blur operation. then when you want an image you run the model backwards.
Yeah, that was my understanding of things too. What I’m curious about is how the Indie Game awards define it. Because if games that use ((Procedural Generation) AND NOT (Generative AI)) are permitted, then that would surely require a way of cleanly delineating between Generative AI and the rest of procedural generation that exists beyond generative AI
most procedural algorithms don’t require training data, for one. they can just be given a seed and run. or rather, the number of weights is so minimal that you can set them by hand.
You can use statistics to estimate a child’s final height by their current height and their parents’ height.
People “train” models by writing a program to randomly make and modify equations, then keep them depending on if new accuracy is higher.
Generative AI can predict what first result on google search or first reply on whatsapp will look like for llms.
There are problems. Training from 94% to 95% accuracy takes exponentially more resources as it doesn’t have some “code” you can fix. Hallucinations will happen.
On the other side, procedural algorithms in games just refer to handwritten algorithms.
For example a programmer may go “well a maze is just multiple, smaller mazes combined.” Then write a program to generate mazes based on that concept.
It’s much cheaper, you don’t need GPU or internet connection to use the algorithm. And if it doesn’t work people can debug it on the spot.
Also it doesn’t require stealing from 100 million people to be usable
(I kinda oversimplified generative AI, modern models may do something entirely different)
From my understanding, AI is the general field of automating logical (“intelligent”) tasks.
Within it, you will find Machine Learning algorithms, the ones that are trained on exemplar data, but also other methods, for instance old text generators based on syntactic rules.
Within Machine Learning, not all methods use Neural Networks, for instance if you have seen cool brake calipers and rocket nozzle designed with AI, I believe those were made with genetic algorithms.
For procedural generation, I assume there is a whole range of methods that can be used:
- Unreal Engine Megaplants seems to contain configurable tree generation algorithms, that’s mostly handcrafted algorithms with maybe some machine learning to find the parameters ranges.
- Motion capture and 3D reconstruction models can be used to build the assets. I don’t believe these rely on stolen artist data.
- Full on image generation models (sora, etc.) to produce assets and textures, these require training on stolen artist data AFAIK (some arrangements were made between some companies but I suspect it’s marginal).
wait it used genAI? It’s coming off my wishlist… :/
deleted by creator
I feel like this is virtue signaling more than actually addressing a real problem with Clair Obscur.
That and attention seeking through controversy
Welcome to the internet. No one knows each other, no one considers context, no one reads past the headline, everyone makes snap judgements based on half understood heuristics, and then rushes to the comments to grandstand. A job that could be trivially done by AI, and almost certainly is, but instead we’ll all pretend like we’re the last bastion of human sanity.
Like that the story is bifurcated and that the combat in the late game is parry or die?
And yet someone completed the game without parrying a single time.
Good for them. Did they do it without dodging too?
Yes. It was reported basically everywhere.
https://gamerant.com/clair-obscur-expedition-33-no-doge-parry-completion-all-hit-run-feat/
Want to move that goalpost again?
My experience was my experience. I’m glad for that person that they found that build. I did not, and I’ll wager most others didn’t either. The last third of my game was spent pumping points into defense and vitality to alleviate the issue, but it was a drop in the bucket. This is like when I vented frustrations with RE2 remake’s scaling difficulty, and someone pulled up, “Well, speedrunners don’t run into this issue, because…” I’m not a speedrunner. I’m a guy playing the game for the first time, and I used the information in front of me to make the best choices I could, and I still came away with criticisms. In CO:E33, it led to situations where the damage was so high and the action economy so constrained that it was faster to throw the fight and reload than it was to take a hit on the first turn and recover from it, and that sucked.
…dawg, I think you’re just dogshit at the game. are these issues you had on the lowest difficulty? Because if so, then no, at no point did you ever “get good”, I promise.
I beat the game on normal difficulty. Believe it or not, you can be good a thing and still dislike it. And I like the game, for the record, but my criticisms of how much weight they give to certain parts of the combat, which changed somewhere around the back half of act 2, mind you, hampered my desire to do more of it in act 3.
What are the odds that a mechanic introduced to you in the first tutorial combat (and continuously iterated on throughout all the prologue combat encounters) is a required component of the game? Crazy.
I think you should probably give up on gaming. Doesn’t seem like your scene: it’s for people that have the ability to process information and learn from it.
I have a criticism or two about one video game, and you leapt to “gaming isn’t for you”.
They’re just incredibly bad criticisms tbh. The first one is only subjectively a “criticism” in the first place, and the other is- at best- a poorly made observation.
So yes, let me double down. If whenever a game says “this is how you play the game” and your response to that is “it shouldn’t be”, maybe gaming isn’t for you.
If whenever a game says “this is how you play the game” and your response to that is “it shouldn’t be”
That response is what a critique is. Metroid Prime 4 says, “you play the game by collecting these green crystals,” and many critics said, “it shouldn’t be.”
…no, that is quite literally not a valid example of criticism. Sorry.
Get good or lower the difficultly and stop crying. Also, you know there’s a dodge button right
Sure do. I got good and still have this criticism.
Yeah.
Maybe a technicality too. The rule said “no AI,” and E33 used AI.
I get their intent: keep AI slop games out. But in hindsight, making the restriction so absolute was probably unwise.
People are saying “it’s fine because it was used in the early stages of the game for placeholder art” but that’s kind of missing the point
The problem is that they used AI and didn’t disclose it, as well as releasing the game with AI textures still in it. Yes, these textures were quickly replaced, but it’s still very concerning they weren’t upfront on how they were using it in the game making process
Edit: there isn’t even a disclosure on their steam page
Horrid article, unless the intention was to throw shit around and hope to cause a commotion. There are no AI assets in Clair Obscur, and it should have been made clear by the article. From the IGA’s own statement:
[…] the use of gen AI art in production […] does disqualify Clair Obscur: Expedition 33 from its nomination. While the assets in question were patched out and it is a wonderful game, it does go against the regulations we have in place.
I think you have missed the actual issue here. The issue is not whether or not the game currently contains AI assets, the issue is whether AI was used during development. Quoting the article (emphasis mine):
“The Indie Game Awards have a hard stance on the use of gen AI throughout the nomination process and during the ceremony itself,” the statement reads. “When it was submitted for consideration, representatives of Sandfall Interactive agreed that no gen AI was used in the development of Clair Obscur: Expedition 33.
“In light of Sandfall Interactive confirming the use of gen AI on the day of the Indie Game Awards 2025 premiere, this does disqualify Clair Obscur: Expedition 33 from its nomination.”
The actual problem is that simply using generative AI during development disqualifies a game from being nominated, and Sandfall Interactive lied and said they did not use gen AI.
The issue is not that the game was disqualified. If the rules clearly and unequivocally state that at no point can generative AI be used (and also clearly state what, in the spectrum from algorithms -> machine learning -> chatbot slop, they consider to be unacceptable, which I don’t know if they did or not, but that’s not the point), then there is no controversy, and I’m not criticising that.
The issue is that the article completely disregards mitigating facts that counter the narrative. There are no credible sources linked in the article save for one that was grossly misrepresented. Critically, we don’t know what Sandfall actually said before the nomination or after, or how the decision to disqualify was made, only the second-hand account in the FAQ. The article presents circumstances in a biased way, leading the reader to interpret it with the assumption that there are AI-generated assets currently in the game. It is, frankly, sloppy journalism.
Do you know where those rules are? I’m genuinly interested in where exactly they draw the line. I constantly see people ranting about gen AI when used for art, but even simple, basic code autocomplete is AI under the hood these days. I can’t imagine developers not using autocomplete.
The rules are on the same page I linked (https://www.indiegameawards.gg/faq), under the “Game Eligibility” tab. I gave them the benefit of doubt and assumed that they had defined the exact terms of what is and isn’t allowed, but apparently I was wrong. Regarding AI, the document contains a grand total of one sentence:
Games developed using generative AI are strictly ineligible for nomination.
I’m assuming the definition of what that entails is “at their discretion”, meaning whatever they feel like at the moment. I see that sentiment reflected in this thread too.
It’s possible that potential nominees have to sign some kind of declaration that they’ve complied with the rules, and that might include a more detailed list of rules, but I have no evidence to support this.
Unfortunately the boundary between “AI” and “not AI” is the polar opposite of sharp and well-defined. I’ve used Allegorithmic Substance Designer a lot for CGI work (before Adobe ate the devs; fuck Adobe, all my homies hate Adobe), and it contains a lot of texture generator algorithms from simple noise to complex grunge textures. Things like Perlin noise and Voronoi diagrams are well-known algorithms and definitely not AI. Chatbot slop is right out, but in between those two, things get remarkably fuzzy and Heisenbergian. What about an algorithm that uses real-world samples, like an image? Or multiple images? Machine learning is not the same as AI, so is that allowed? Where’s the line? I’m reasonably certain that everybody has a different answer for different situations based on different criteria.
So any game whose developer has used a recent version of VSCode will be disqualified in the future? VSCode has a GenAI autocomplete turned on by default.
One single question about an API to ChatGPT and your game is out.
Use Photoshops generative features for a marketing asset: out.
You get how insane the rule is?
You can only qualify it you write your game in vanilla vim with no extensions and graphics must be drawn in an old version of Gimp? 😆
So insane that people have to go back to the primitive workflows of… 2021 🤣
The problem is that Chinese, Indian and Turkish developers couldn’t care less about western AI purity tests and will blast past any competition who does.
Unless Captain AI Planet stops them, the cat is out of the bag and not going back.
If you want to run a AAA live service game the amount of content you put out every X weeks is how you make money. And the one who can keep up the best amount/quality ratio will always win.
The average gamer won’t care if the latest Gooner F2P or FPS game DLC is AI generated, AI assisted or lovingly hand crafted. They’ll throw their money at it anyway…
I was just about to buy this game too
Not now.
Fuck AI.
They used it for background assets but patched it quickly. It’s not as egregious as it sounds.
Read the article. I understand.
It’s not an excuse. I just simply don’t support AI slop. AT. ALL. I don’t care that they used it for a placeholder texture. And that texture didn’t end up in final release. If I have anything to say about it, these companies will not get my money if they dipped their toes in AI. AT. ALL.
There are workers that could have made placeholder textures for fucks sake.
There were no artists who would have been paid to create placeholder textures. They would have used textures from open source sources, or from previously used textures.
I agree that AI should never replace the job a person could do, but in this instance the user of AI caused no loss of income.
AI has been around for years well before the LLM craze. It’s not all evil ffs.
A contractor used it for placeholder background assets in 2022 when the best genAI tech available was DallE-Mini and literally no one was even remotely aware of any ethical concerns yet.
Like I get that time feels nonexistent since 2020 but people seem to forget genAI was literally not even a known thing until 2024 at the earliest.
Today genAI and all those who support it in any capacity deserve to rot in hell, and it shouldn’t even be used for placeholders or concepts at any stage in the process. That’s a given. But why are we applying 2025 standards to something clearly done at a time before anyone knew anything about this shit?
Also the article people keep bringing up about Sandfall admitting they use “AI” is nothing but conflated buzzwords and has literally nothing to do with genAI, they’re very obviously referring to regular ass machine learning and Unreal tools that have existed for ages.
lol dumb disqualification. Game is overrated anyway
Insightful commentary, I’m glad you shared your thoughts with us.
Triggered over that comment, get real lmao. It is very overrated
Wow, two bangers in a row. How do you do it?
AI isnt needed at all, we didnt need it in the past to create art. And with all the tools and knowledge available online, for free, theres even less reason we need it these days.
I’ve never pirated a game, but if developers are going to use pirated content to make a game, they cant be mad when we pirate their game.
Seems a few “artists” disagree with you.
“make a dirt texture”
Being a 500 person studio with a 400 million dollar publisher means you still qualify for the Indie™ Game Awards but using ChatGPT to make a random powerpoint is just a bridge too far.
Apparently Blue Archive, the game that was given the award after they disqualified E33, ALSO used AI.
I heartily approve.
Because so many people are blowing up without reading the article I felt it was worth posting this. Based on the wording it sounds like they were not disqualified for having AI in the game, they were disqualified for not disclosing AI had been used in development.
“The Indie Game Awards have a hard stance on the use of gen AI throughout the nomination process and during the ceremony itself,” the statement reads. “When it was submitted for consideration, representatives of Sandfall Interactive agreed that no gen AI was used in the development of Clair Obscur: Expedition 33. “In light of Sandfall Interactive confirming the use of gen AI on the day of the Indie Game Awards 2025 premiere, this does disqualify Clair Obscur: Expedition 33 from its nomination.”
Additionally, here is another article where they are clarifying HOW it was used.
Following the publication of this article, Sandfall Interactive wishes to provide the following clarifications. The studio states that it was in contact with El País on April 25 - three months prior to this publication. During these exchanges, Sandfall Interactive indicated that it had used a limited number of pre-existing assets, notably 3D assets sourced from the Unreal Engine Marketplace. None of these assets were created using artificial intelligence. Sandfall Interactive further clarifies that there are no generative Al-created assets in the game. When the first Al tools became available in 2022, some members of the team briefly experimented with them to generate temporary placeholder textures. Upon release, instances of a placeholder texture were removed within 5 days to be replaced with the correct textures that had always been intended for release, but were missed during the Quality Assurance process.
TL;DR: They experimented with Generative AI when it first came out, used some of the results as temporary assets that were always intended to be temporary. They still got in to the final product because QA missed them, which was promptly fixed in a patch. Indie Game Awards disqualified them for failing to disclose this in the first place.
Key takeaways:
- AI didn’t steal anyone’s job in this instance. It was simply used as a tool to help make an artists job easier.
- It was never meant to be a part of the final product, and currently isn’t.
- They used generative AI around when it when it first came out, probably before most people started realizing it was being trained off stolen artwork as well as a lot of the other problems with AI.
Make of that what you will. I personally think this is being blown out of proportion. They made a mistake and have openly corrected themselves. Good for them.
