AI generated models flooding the website

Just to add to this conversation.

I’m not against using AO to generate ideas, unlock creativity, etc…
The problem is, and using Deltaprints example, is the inability to deliver as a 3D model what the picture shows. And by the time you get to know this, the download is already done, contributing to the stats!

So the only think to try to get makerworld attention is to DO NOT DOWNLOAD, search for alterantves…and REPORT, REPORT, REPORT…but in order to work this needs to be done massivaly by EVERYONE interested in get, fair true high quality models and be rewarded by that!

1 Like

Most often you can see glaring differences in the Print Preview of the profile, compare to the suspect image and particularly in figurines the crisp edges and features drop off. Takes a minute and you can usually work out if the model will work for you, if not move on and download something else. For most in this thread we’re able to raise something printable anyhow so its kinda moot :slight_smile:

1 Like

True,
O look at the 3d preview is enough for experienced 3d printers!

The 3D printer scene is currently being flooded with beginners.

2 Likes

That is not necesseraly a bad thing. What we need as a community is to help, guide and correct bad behaviour. There will be of course someone who will not understand or refuse to comply…

So… to conclude this topic: where do I generate my AI models? :smiley:

2 Likes

AI generated picture no, Fusion handmade model yes

Sure it cannot be a direct copy on every single parameter, however the design idea itself is AI generated so everyone can use it

2 Likes

I don’t know if that’s so beneficial if Makerworld at some point consists of the same models and design ideas that were recreated by 500 different creators. Well, let’s see where the path goes.

I agree, that’s why I don’t think having AI coming up with the design is the way forward :slight_smile:

3 Likes

Without starting a war or anything like that, since it’s something i am trying to avoid for a long time.

But you say:

And yet, most of your generated images consist of already laid down models by other creators. After all, the AI is trained on those same models, available on the internet. So, if we are going into the rabbit hole, none of your models are unique in any way possible since they were recreated from other existing models. This is how LLM’s work. They get trained on available data.

You can ask, does he know what he’s talking about? And the answer is yes. I run not one but several local LLM’s and even a couple trained for me, with my data of interest. ComfyUI is very powerful and the data models can be trained on art only, there are lots of data models already available. I could explain how all of this works in detail, but… AI can teach you that too right?

Keeping the point of copyright, again, there is no copyright for AI generations. Period! Creating something in fusion taking inspiration from other designs is what makes a real creator, not the other way around.

And to finish it off, and to be honest, 99% of your “models” are not unique and they already exist and have existed before you tried to replicate it through your images. I could go way back in time to search for examples and post them here but it would be of no use because… no one creates what you create, from what you write every time.

Look, my coasters are really easy to make. There are tons of them already. Should i stop making them just because there are already similar models? No, of course not.
My keychain fidgets are pretty simple, yet there are lots of variations of mine already and i created mine from scratch way before. Should i go and report people who invest their time and even make better versions just because they got inspired by mine? No! This is what defined 3D printing and it still is what defines it. Real creators understand that. Do you?

You try and replicate from images. It’s ok, not everyone has their own ideas, i get it. Others try and improve already existing models and others even have their own ideas and put them into 3d models. It takes time, it’s work put into it… I like to make fun stuff or, like it happens a lot, i look at something in my garage and think “hey, i could do a model of X to hold that thing in place” and through iterations (yes, i do lots of prototyping so) and lots of fails and wasted filament, come up with a working model for my needs. No one will need what i created right?! But why not share it just in case? That’s what i do.

Either way, i am sick of this discussion, especially with you. I know, you post an image of all your likes and download count and wow, congrats! In the end, i make what makes me happy and you too, it seems. Let’s keep it this way as i already have said. After all, this topic is not about you, it’s about everyone that misuses AI as a whole.

Have a great day, like always :slight_smile: :ok_hand:

Hi, yes you are right. But since it’s okay according to the guidelines, new creators will always come up with AI-generated preview images. we see now. So far you can’t do anything about it. As many discussions as you have here, as many reports as you produce. It doesn’t matter how many fake accounts you use to spread bad comments. I think it revitalizes the rapidly growing 3D model world. Everyone should decide for themselves whether they are for or against. Just like it was said above. When I download something, I look at the 3D preview. The other pictures etc. and then I decide to print or not. But, to each his own.

Since there’s a discussion about end-result, I’ve made a simple tutorial on how to check whether a design is a direct copy or a remix or original: Point farming profiles - #63 by Johnny_Bit - I’ll say with certainty that there’s virtually no overlap between one moon design and another, the median distance is too great.

As for taking inspiration from image, be it ai generated or actual - there’s no problem there. Hell, I’ve even got inspiration from real life object I had in my hands and created a model from that inspiration. However, the problem with the AI-generated covers is that they do not represent actual object. Personally I’m sorta-ok with renders, but ai generated cover image that has no relation to actual object except for being inspiration for it is straight up lying. Instead of using ai-generated image for cover, take a good photo of the object! Here’s a great resource thread on how to make a decent photo: The Ultimate MakerWorld Photography Guide - I recommend that to anybody! I probably should take more from that…

3 Likes

Well, I’m a professional programmer and I think AI will change our lives at some point, maybe soon. Just as humanity is programmed to one day destroy itself. In my opinion, the latter is only a matter of time, and yet it will not change. Each individual decides whether he will stand idly by, resist unsuccessfully or use the developments to his advantage during his lifetime.

AI-generated models are apparently also being uploaded elsewhere. Cults is also affected by this. Just one example:




I agree that AI is here to stay, and I welcome it, I use it daily on my day job, but I’ve said it before and say it again, I use AI to tackle low value added activities so that I gain time to think and create, I don’t need AI to replace me on that, I need it to replace me on menial or repetitive tasks (which are not always simple and it does require a strong model to output proper results)

3 Likes

This is true also and should be a part of this discussion. But i must ask. Does that image correspond to the model? If yes, then the creator did a nice job. If not, then it’s like it was said before. The printed model should be 99% as good as the image.

AI is a wonderful thing if applied correctly. It can also be a tool for the lazy. That is the way i think. AI can only create what was “feed”, in this case, it crawls the internet for models and categorizes vertices, meshes, file names, etc., and then it throws it at the user that asked for it, for example: “create a 3d printed model, with visible layer lines of an owl, very detailed, with intricate design choices” and the model will throw an image of that exactly. Creating said model after is another thing and that’s where the hard work has to be done. And to be honest, not everyone can get to that point.

But yes, i get your point and it was a good point indeed :ok_hand:

No, the model is available without colour. The STL for it is colourless. But it was created with the second generation of an AI, and it looks very good.

But since we are here on Makerworld:

This is what it looks like from another perspective:

And so as STL:





1 Like

Thank you. Going to see if i can find that model then and try and print it only in 1 color to see it live. I am always eager to find how AI is evolving since i follow this since it came out (i have almost 8Tb of AI trained models in my hard drives) :laughing:

Ultimately this will always be a passionate topic as the purist sees the skills and knowledge as well as the time that went into a piece and just like good art it has layers. Sorry but AI is cold and sterile, it has instant appeal to the masses as “sex sells” and we’re often fooled by a pretty picture and never want to see it without the filters.

This thread won’t solve the problem, basically MW is happily pushing the AI trolley down the aisle and dealing out printmons etc to sell printers and accessories. We don’t figure in this equation and I’m sad that we’re not rewarding quality over quantity.

2 Likes

I like to use ChatGPT to design circuits, not models. The reason for this is I am afraid AI may accidentally plagiarize somebody else’s work. Also, AI models often turn out to be junk in the end. You can use AI to find or give you instructions to make a model, which isn’t “bad”. But having AI do all of the work for you is “bad”. So in conclusion, AI is a great tool for Makerworld, but we shouldn’t abuse it.

2 Likes