Internal test game created by a single (senior) developer in 5 days using our new tools and engine. Uses WASM mini-engine (ECS, rendering, animations and physics). 100% AI generated assets using our new asset pipeline. Custom character controller. Slop or not?

All you've done is accelerate asset flipping minus the provenance. It's absolutely slop.

I have not been part of the game project. It's a single developer project.

Okay, swap out "you've" for "they've". That was difficult.

@SebAaltonen don't worry reid, buddy used to work for ubisoft and unity. he knows and is deflecting dispite this being his work being utilized for the demo.

@Rahll Yes, this project uses mini-engine developed by me and our team. The game and content is not made by the engine team. You can criticize both if you want. What are the technical/engine issues you see in the above video?

@0JMachine "100% AI generated assets using our new asset pipeline." Are you generating the assets for them or not?

well, there is nothing even slightly artistically interesting about it...

I see a lot of AI slop games in my feed daily. This looks way better. Just my opinion of course. I am biased since it's made by a person I know.

I'm sure @Nintendo have zero problems with this. 😂

@Nintendo Definitely have to be careful with AI asset creation if you intend to ship a game. These things might come from the AI training set or from the reference images you use in your art design. Human artists are more careful with reference art. Don't copy details like that.

@Nintendo Ah, yes, there is absolutely nothing that could go wrong when you have thousands of assets, each of which could infringe on the IP of any of hundreds of thousands original sources, not just in games, but pretty much *everything* protected as intellectual property.

@SebAaltonen @Nintendo You have no idea how many AAA games before AI committed plagiarism or used STOLEN assets from other projects and other IPs, etc. This is what happens when an idiot comments without having worked in the industry.

@SebAaltonen @Nintendo In your opinion which are the three most notable examples of this happening?

What’s the game? Always see people say game and it’s just walking/driving around with no goal

Game starts with a good character controller. First you make the main character locomotion feel good, then you build more environment. Would you expect to see more in 5 days by one dev?

Not knocking the technical side. Which undoubtedly is top notch. Game implies some sort of rule set, goals or obstacle to overcome. A chessboard with no pieces or rules is hardly a game. Just being pedantic

Shows clearly the difference between a senior gamedev versus random AI vibe coders with no gamedev exp. We have way better AI tooling too, written by professionals of respective gamedev areas. AI generated meshes need to be remeshed to be usable, etc.

I have told the developer that there was art resembling the Zelda emblem in the pillar. That asset will be regenerated. Thanks for noting me about it! We also started discussing about automated tooling for detecting potential copyright infringement. Many AI models have some internal detection, but it's not perfect. Not good enough. We need to implement tooling on top to ensure devs can safely ship AI generated assets.

Quick AI copyright check: Finds the issue instantly. If you intend to ship AI generated content, better to validate it. The above video is an internal character controller proto/template. Not a shipping game. If you intend to ship, check for copyright infringement.

I want to apologize for not screening this internal dev video properly before I posted it. Next time when I post videos with AI generated assets, I will check thoroughly that there's no dubious content. Thanks for people pointing out the crest logo. I didn't notice it.

Improved version without the infringing object and new features:

The Triforce looks sloppy

Yeah. Unless it's a Triforce-inspired rune :D

@Plebian_2 What is this stupid ass reply, you've already been exposed 🙄

@Plebian_2

Not slop. Not a game. Just an "engine example".

There's no character system in the engine at all. No walking, no climbing, no swimming. Nothing. All of that is custom game code. The character controller is the most important part of any game. If it feels good to control your character, you have a very good starting point for content production.

Yes, sure. Most of WebGL engines don't have any character controller as well. But they do have examples that implement one. Just like on the vid. Less visually appealing though, as this example above has more polish.

5 days is crazy when the entire thing you show is a character walking and climbing around on assets that took no dev time to make, plus the animations look jank

I have seen much worse results done in much longer time. This is not Unreal Engine or Unity or similar. There's zero character specific code in mini-engine. Do you write a similar character system from scratch faster? Does it feel good to play?

not personally im ass at game dev, but solo game jams exist and alot of devs can make games alot better than this slop entirely by hand and sometimes in a shorter timeframe

@SebAaltonen You lot with Japanese anime profile pics are nothing but talk.

@SebAaltonen just because I cant do it myself doesn't mean that nobody can fuckin duh

Appreciate that I can now just mentally tune out anything that looks like this now.

having a good dev is half the battle won already. assets can easily be found online, AI is redundant

Online assets are bad because they don't fit your art design. And you have to mix multiple asset packs created by different people. Hard to make cohesive looks this way. Better to write a good art design -> generate concept art -> generate unique assets for your game.

sure it would look uniform and work together but it's pretty questionable how much say you have on the final model when you're generating them as opposed to modelling?

There's quite a lot of control with modern image to 3d models. You iterate on the image rapidly. You can even paint on top if you want that. Then you generate 3d.

that sounds handy, still an artist would be the best choice if you can afford it

You'd assume a "senior" dev would recognize the triforce. AI typers don't even play games and yet want to be devs. Classic, wanting the reward without the work lol

the senior dev wrote an ECS, a renderer, a physics engine and a character controller in 5 days the AI made some textures guess which one the post leads with

The mini-engine WASM library was written in 4 months by a team of seniors. It's used in several small in-house games. AI made all the 3d-meshes, rigged the character, made textures, terrain assets, grass, skybox, set up the renderer, engine and physics and wrote the character controller and input handling. And of course all the web app code. Mini-engine is just a tiny WASM library. Not a framework.

As much as I'm against AI-generated assets on humane principles, this is impressive* as fuck. Probably looks worse without twitter compression, but it's orders of magnitude better than any purely vibecoded slop I've seen here. *Talking about visibility of AI artifacts. Your remesh is clearly working. Artistically it's still slop haha

Slop because of the breath of the wild climbing but not slop otherwise

You don't like climbing like that?

It was mostly a joke. Sorry lol.

It's also plagiarized slop, because if that isn't a barely distorted version of the Hyrule crest from the Zelda games, I'll eat my hat. @NintendoAmerica #plagiarism

What’s the point? You’ve created something that’s maybe 25% as good as what someone could throw together in Unreal engine in a few hours. Is the market really dying for worse versions of things that already exist?

is the music AI-generated too? if so I think a certain Japanese corporation might want a word

Music was just slapped on for the video. Not part of the game.

was “Zelda” part of the prompt used somewhere? the design on that column looks very very similar to the Zelda royal crest symbol

I didn't ask about his prompts. He is a Zelda fan, so it's possible that he asked for a similar art style, or had that kind of concept art in his art design.

Slop. Lol All U got is aswd basic controls and ripped Zelda props

The most "impressive" here is the custom climbing. A custom engine is always special but why is it better to use than Unreal or Unity? For the 3d models you can get better ones via asset packs, the AI ones look a bit ugly and strange imo and the column looks like a Zelda rip off.

Asset packs don't have cohesive art direction. That's the problem with combining multiple packs from different authors. Better to spend time on good art direction document and generate cohesive unique assets for your game.

I'm not against other devs who use it AI to just prototype (not the final game assets!) but you should look out more for "ripoff" content (Zelda column). Can you tell me why the custom engine? Is it a requirement for the access to your pipeline?Why use it and not Unity or Unreal?

took 5 days for putting some assets and adding a climbing anim?

5 days includes creating all the assets: terrain, grass, sky, props, character, animations (+rig) and the character controller (run, walk, climb, swim). There's no character controller in the engine. And of course tweaking the character controller to feel snappy and fun to play.

7 seconds in and you're sued by Nintendo. lmao. That's the problem with using AI to generate assets stolen from other artists.

5 days ?? .. thats shockingly long and no .. no sarcasm.

I have seen a lot of slop that looked way worse after months of development. And played much worse too.

I'm not a programmer (though I used to work in 3DS Max). But I figured that in a game engine, you just create a space, add some grass, put some water in it, add an avatar... and you're done. An hour's work? Fifteen minutes?

You should try it sometimes. You can do something in 15 minutes, but it wouldn't look and play like this. The main thing about this video is the custom character controller with walk, run, climb and swim controls. According to the author, most of the time was spend polishing the character controller. Also as you know from 3DS Max that character modeling, animations, rigging and skinning take quite a bit longer than 15 minutes. The animation blending is quite simple in this proto (to make transitions feel ultra-snappy). Not much time spent in animation state machine fine-tuning.

In 2026, I shouldn’t have to spend 5D agonizing over totally trivial things that should have been routinely handled for over a decade. Game dev is in a massive crisis cause production is extremely slow, inefficient, and the huge CPU/GPU/MEM/SSD perf is incredibly underutilized.

not sure assets are even where slop shows up first. the custom character controller feel is usually the real tell. how much of that 5 days was manual tuning on that vs letting the pipeline run?

Of course there's a lot of manual tuning. That's what game development is all about. If you don't iterative, you don't get good quality. Only a human can tell what fells good for human to play. There's no way to avoid playtesting. And I don't think anybody should even try to build a game if they don't like playtesting it themselves.

"AI copied"

Not bad. The question is though, are those AI-generated assets optimized enough in terms of vertex/polygon count? Also, what about LoDs? Is your AI-assisted asset generation pipeline capable of producing multiple LoD levels of the same asset in a deterministic way?

Raw AI generated assets from all the generators we have tried are horrible. Thousands of tiny UV islands. Doesn't LOD well at all. We remesh AI source assets to improve the topology. Then we reproject the high mesh geometric details on medium poly asset and bake new normal maps. This has ~10x less UV islands so it LODs much better and the starting point has 10x+ less triangles, but looks almost identical as all fine details are preserved in the normal map. This pipeline is fully automated. @JarkkoPFC posted some examples.

Yep, expected about as much from gen AI-driven game asset generators lol. But yeah your solution sounds quite efficient, and if it works well enough for your needs, then it deserves commendation!

u wanna partner to deploy this type of stuff (multiplayer) on ? ai cloud coming soon, i love video games

just embrace the slop. turns out humans like slop

it's definitionally slop but also it seems quite obvious visually

Thank you for highlighting one of the major problems with AI: you can never tell what it's ripping off when you use it. I mean, unless you intentionally stole this symbol from the Zelda games.

Prototyping for gameplay and code and whatnot is neat but I would never let AI anywhere near art assets whatsoever. Here it looks fine but not very cohesive. Cool demo in terms of tech otherwise

Depends on your quality standards. Roblox games look way less cohesive as most people don't create custom art at all. They download assets from asset store. Same problem in many indie Unity games. With AI, you write a art design document with reference art and iterate on it. Then you generate cohesive assets using the same design. Not as good as having a team of professional artists in-house, but most indies can't afford that. AI is a good alternative way of getting unique assets for your game that follow your art style.

@Dispatch_Graph While I 100% agree that having an artistic vision and direction is key to make your game standout and have a unique/recognizable style, I find the argument that "we can't afford artists" to be quite dubious. From experience they're already the least paid people on the industry.

Indie teams are often <10 people. Couple of artists. There's not much throughput. It makes perfect sense to focus the art effort for creating fantastic art style and all the hero assets and AI generating all the props and similar. If you get your props from asset store, they are not cohesive with your unique art design.

That’s some slop.

the fidelity of this given current AI capabilities says the engine is either too minimal, requiring too much upfront knowledge or has too many rails impressive demo but not very convincing tbh

Nintendo seeing the Temu Hyrulian Crest.

This reminds me of Dreams.. looks cool!

The real question is , will the price drop dramatically? Because no way in hell I'm paying 70 bucks for AI made games.

Fully AI made games will never be successful. You need to be a gamer to understand games and have the passion for game design and playing. Game dev is all about iteration and testing repeatedly. If you don't like testing your game and don't love that particular genre, you will never make anything worth playing. But even devs with passion should use AI for repeated and low importance work. Both in coding and in art. Of course you need at least one artist in the team to define what the game should look like, do concept art and make initial set of assets. AI can do more variants and create less important props and build structures, build terrain (rocks, trees, vegetation...), less important NPC dialog/voice, etc. We humans should focus on what matters: hero assets, bosses, story, making the character feel good to control, game/art design, etc. Same is happing in programming too. Not every piece of code is equally important.

I understand what you're saying and I wholeheartedly agree, yet I'm worried about the future, think of all the human made slop that somehow sells, if the AI slop starts gaining traction, that's it, gaming as we know it, is done.