Forum adverts like this one are shown to any user who is not logged in. Join us by filling out a tiny 3 field form and you will get your own, free, dakka user account which gives a good range of benefits to you: No adverts like this in the forums anymore. Times and dates in your local timezone. Full tracking of what you have read so you can skip to your first unread post, easily see what has changed since you last logged in, and easily see what is new at a glance. Email notifications for threads you want to watch closely. Being a part of the oldest wargaming community on the net. If you are already a member then feel free to login now.

Someone tell me I'm being paranoid.

Please look at the following pictures.



And it's just scrolling through the first two pages sorted by new. No, this is not the same creator. It's 6 different ones, every one with at least a dozen similar Kickstarters as their track record, each campaign has nearly the same description, all are super short campaigns, similar render style. The models look AI generated at that. Cash grab scam?

Cash grab? Sure.
Scam? Depends on if they deliver or not.

Does KS have any rules against AI junk?

I saw these, or something like them, whilst watching a twitch stream a few weeks ago and they were discussing them then. The very fact that some of the models have impossibly spindly whiskers sold me on them being AI generated junk. No human would design them that way.

Are things like these popular?

I'm twitching imagining how many supports those damn things will need. Especially the dog ears

Yeah, they look AI generated, but AI can make STLs, so as long as the STLs are viable (and I bet they aren't test printing them first) and they actually deliver I say have at it.

If I wasn't so completely sure they're unprintable I'd go for the space doggos at least

Charax wrote:
I'd go for the space doggos at least

The most terrible slop of the bunch. Curious.

JWBS wrote:

Charax wrote:
I'd go for the space doggos at least

The most terrible slop of the bunch. Curious.

Hmm, chapter apropriate dog heads for power armour could be a thing. Im calling Golden Retriever as Blood Angels,

Oh, wow. Those campaigns are so shady. All 'different' creators with similar looking campaigns, down to the phrases used and campaign layouts? Plus super low funding goals, barely any updates (usually just a 'campaign is fulfilled') just smells of something being off.

Also interesting how you can see when these creators (still my opinion they're the same) switched how they did their campaign thumbnails to the new style that was shared above. Also fascinating how most of the previous campaigns were in pesos, but then switched to USD. Another interesting point is saying the business is located in California while some of the creators say they're in Mexico.

Best part was I kept trying to find out if SOMEONE was displeased with their pledge and said something. If these -are- poorly made, you'd think someone would call them out. Looks like most people buy and forget, but did find a comment here (https://www.kickstarter.com/projects/outterspace-design/royal-guard-dogs/comments) that was exactly what we expected to see. Supposedly poor craftsmanship, clearly not designed to print, and potentially done with AI. However no AI disclosure and potentially multiple shell creators that are just one person....I'll say it again, but wow.

The whiskers in the first two are a dead AI giveaway: the change color based on the background /other figures uniform. Wouldn't trust a campaign using AI images.

There are a few AI STL tools around, usually picture to STL, which would work here. Generate the image or a few iterations in one of the AI image tools, upload to the STL tool, call it done.
Here's one for example: https://www.meshy.ai/

Go to community and it pretty much matches the comment from the post above. Basic shapes, missing a lot of the details the painted models have.

Sadly, there have already been 100% AI-generated kickstarters, massively popular too.

This one made 64 thousand euros for typing some words into a box then making up some nonsense about it being really hard work because they had to fix the files afterwards to make them printable:

https://www.kickstarter.com/projects/86900743/1001-fantasy-bits-3d-printable-stl

The market, broadly speaking, does open their wallets for this stuff and doesn't care all that much about whether AI is used or not.

I work as a sculptor and would say I've probably got 5 years left before the entire market is millions of auto-generated AI product listings.

I think I'll probably switch to hand sculpting and 3d scanning since that will be one of the few ways to prove legitimacy as a sculptor given time, in the same way that some 2d artists are switching towards traditional physical media.

 Dysartes wrote:
Cash grab? Sure.
Does KS have any rules against AI junk?

None whatsoever, other than having to declare it way at the bottom of the project buried in the FAQ.

Cults 3D has strict no-AI rules.
MMF has a site-wide ban on scraping the site for AI.
Thingiverse filters AI models.

Overall, the situation and support from industry is better than most other places (probably because sites don't want low-effort garbage flooding their platforms), but kickstarter and patreon are still the wild west where anything goes.

From five months ago...

"I like these designs, and I've already backed Pumpkin Fellowship, Medieval Town, Steampunk City, and Fantasy Town.

I suspect you all are the same people behind these several studios, am I right? If so, you're not really doing things in 75mm scale; You're merely putting all designs at max 75mm height which is not the same, since the difference is astounding between the models sometimes..

Also, why are you guys releasing similar design innovations through different subsidiary studios?

https://www.kickstarter.com/projects/1214288519/elven-buildings/comments?ref=creator_tab

Buyer beware, I suppose. Look into previous projects and read the comments.

scarletsquig wrote:
5 years left

Nowhere near 5 years, except perhaps for the most talented few percent of creators.

I noticed something fishy going on, specifically with two projects by two differemt companies giving the exact same response to the exact same person with the question(can I print them without bases). Another fun fact: two of the companies opened their accounts around early 24. Possibly a coincidence, but I havemt checked the other ones.

Sadly, I'm thinking about hitting Gamefound from now on: they at least let you report a project if it looks lile a scam...

Ooooof. Yeah those are rough. I'm not 100% anti-AI (I think machine learning/procedural generation as a technology is fascinating and has legitimate uses that are being strangled by the "press button, get product" grifters, with discourse on both sides being ruined as a result) but this is just weak.

What's doubly pathetic is that AI is actually quite capable of making far more convincing fake mini pictures than that. I messed about with one of them (ImageFX?) and got this, which looks like an actual model.

scarletsquig wrote:

I work as a sculptor and would say I've probably got 5 years left before the entire market is millions of auto-generated AI product listings.

I think I'll probably switch to hand sculpting and 3d scanning since that will be one of the few ways to prove legitimacy as a sculptor given time, in the same way that some 2d artists are switching towards traditional physical media.

Relying on legitimacy is a fool's game. Customers care about two things, price point and quality. And the second is where you win. You can't beat a double click effort. But quality? There, you still have the ability to retain the lead in a way AI can't replace.

To put it bluntly, printability is something plenty of existing sculptors get wrong. Broken meshes. Voids. Details too fine for the size of the print. Scale completely off compared to other models. All these problems and more exist with files from lazy sculptors - and they're vastly magnified by AI. A computer cannot measure an existing GW kit and ensure a boltgun has 100% compatibility with their Marine arms. A computer cannot read what type/resolution of 3d printer is being worked to and tailor accordingly. A computer cannot consciously design a sculpt so that when it 3d prints, it naturally layers upwards at a 45 degree angle so as to ensure easy and effective printing.

All these 'image to files' are already pretty much as good as they're going to get. The tech itself has no further way to improve. It can already draw an outline in 3d space of a 2d image. But that's it. There's no further data to draw from. They can refine the model as much as they like, but it can't do the sorts of things I'm describing above. And never will be able to.

For the crap sculptors who don't know anything about 3d printing and make unusable hunks of junk though - they might struggle a tad.

Quality or even printability makes no difference... not even legitimacy.

If digital sculpting/printing has proven anything to the miniature industry is...its a race to the absolute lowest prices and maximum quantity of files on a PC drive... to never be printed.

Meanwhile these Ai peeps make money and will make even more with huge flooding of files, to the point someone thinking in creating producing something will not even bother.
Thats basically what digital did to traditional and now this is digital on steroids.

If people want this and invest in this who are we to complain right?

> I work as a sculptor and would say I've probably got 5 years left before the entire market is millions of auto-generated AI product listings.

Probably said that about digital sculpting when it became popular. I presume digital sculpting has taken over everything plastic? Does Reaper Bones still have hand sculpted models?

AI plus 3D printing will let consumers create their own miniatures, not that HeroForge and other digital sculpting software doesn't do that already to some extent.

EDIT: I'm also going to guess that, for larger companies (eg. CMON, FFG) that sculptors (and artists) are not a major cost for a project where the manufacturer casts the miniatures, so AI savings isn't as important to them. However, if FFG can use AI for playtesting...

EDIT EDIT I don't get crowdfunding for 3D prints when there's no manufacturer involved who wants a large deposit, or videogames when they sell at reduced prices after a period of time. With manufactured mini's (and expensive boardgames), it's FOMO, since miniatures can go OOP, game stores won't carry expensive games, or, like Archon, retail *is* more expensive and crowdfunding the better price. (If it's a crowdfunded retail boardgame without excessive miniatures, you can always wait for a sale, unless you need it NOW NOW NOW.

If it looks like BS, Walks like BS, and talks like BS, guess what?

They are all cash grabs and each and every one is a scam. Either Chinese gold farmer style, or just outright nefarious.

Avoid at all costs, no matter the kitten eyes, and soft looking figures approach. Go with it all being a lie. Kickstarter for gaming is dead and dying. That craze ran it's course just because of hive scum like this.

 ced1106 wrote:
> I work as a sculptor and would say I've probably got 5 years left before the entire market is millions of auto-generated AI product listings.

Probably said that about digital sculpting when it became popular. I presume digital sculpting has taken over everything plastic? Does Reaper Bones still have hand sculpted models?

Digital sculpting is faster than hand sculpting especially when you can re-use components. But AI pretty much removes all barriers: time, ability, viability, sanity, etc. Anyone can crap out a million derivitive and unprintable minis and flood a market place in less time than a single good miniature can be sculpted digitally or by hand.

I don't think the problem will be quality as such - most people will be able to see it's garbage and skip it or demand a refund, but given the cost of entry is so low it may become virtually impossible to find anything good. With automated account creation on platforms you can't even just block the bad suppliers because each new wave of crap can come from a different account.

Herzlos wrote:
Digital sculpting is faster than hand sculpting especially when you can re-use components.

Non-digital sculptors also re-use components. I don't think the time difference is actually that great for good sculptors... the medium is different, but the process is largely the same.

It's definitely faster to turn out masses of poor quality sculpts digitally, though.

 Ketara wrote:
All these 'image to files' are already pretty much as good as they're going to get. The tech itself has no further way to improve. It can already draw an outline in 3d space of a 2d image. But that's it. There's no further data to draw from. They can refine the model as much as they like, but it can't do the sorts of things I'm describing above. And never will be able to.

I don't think this is true. Slicer software can already analyse an stl and tell you where it's likely to run into print issues, and in some cases can automatically correct those issues, although the effectiveness of those corrections varies. Some sculpting software can do the same, analysing a sculpt for suitability for various production methods. So building similar capabilities into an AI 3D generator definitely seems like something that should be possible - generate the mesh, analyse it for potential print issues, automatically correct those issues by adjusting the mesh accordingly.

It's a terrible direction for the industry to go, for all sorts of reasons... but does seem like it's going to happen sooner or later.

> unprintable

Good point. IIRC, In early digital, you had these nice looking digital sculpts with tiny spikes and other unprintable bits, even though they were designed by a "sculptor". Eventually, we got to the point where digital sculptors would make miniatures that could be manufactured.

So, to repeat the wheel (: I'm seeing several tiers of AI being used, at least in the near future:

* Companies: I suspect AI's going to turn artists into directors and editors, like it does with video and pictures. At least for now, you'll need someone with digital sculpting experience to test-print or otherwise evaluate and "edit" a sculpt, not just for 3D printing at home, but also for miniature manufacturing.

* DIY: Like AI pics and videos, you can make your own miniature, but, as you said, whether or not it's printable may be something entirely different. Again, I suspect we'll have something like HeroForge, where additional "rules" will be added to make printable miniatures.

* Buyer Beware: AI pics shown instead of actual prints. Sorta like the days of KS when you could show off a nice design but didn't need a working prototype!

I figure most buyers, at least after being burnt will only buy digital sculpts, AI or otherwise, from reputable sellers. Myself, I buy cast miniatures, and wouldn't know if CMON or Reaper was making them with or without AI.

 ced1106 wrote:
* Buyer Beware: AI pics shown instead of actual prints. Sorta like the days of KS when you could show off a nice design but didn't need a working prototype! .

It's not even just KS. There are an awful lot of sculptors out there only showing digital renders of their models. It's a massive red flag, for me.

 insaniak wrote:

 ced1106 wrote:
* Buyer Beware: AI pics shown instead of actual prints. Sorta like the days of KS when you could show off a nice design but didn't need a working prototype! .

It's not even just KS. There are an awful lot of sculptors out there only showing digital renders of their models. It's a massive red flag, for me.

To be absolutely fair, actual 3D renders of legit sculpts are leagues better than AI-genned pics. Obviously physical print pics are the best (I'm particularly fond of the stock-images for Black Cat Cartel's tanks, which often get fully painted prints showcased allong with renders) but I'm willing to risk an .stl that's marketed with renders, whereas one that only shows AI genned pics is right out.

The problem is that it's going to be increasingly difficult to tell the difference as AI improves. Heck, if the pictures at the start of this thread were of individual models rather than groups of almost-the-same-but-inconsistently-designed quasi-clones, most would pass for badly designed digital sculpts quite successfully.

But even with actual, human-made digital sculpts, if the sculptor doesn't care enough to check that the model actually prints, I'm not going to throw money at them and hope for the best.

 insaniak wrote:

I don't think this is true. Slicer software can already analyse an stl and tell you where it's likely to run into print issues, and in some cases can automatically correct those issues, although the effectiveness of those corrections varies. Some sculpting software can do the same, analysing a sculpt for suitability for various production methods. So building similar capabilities into an AI 3D generator definitely seems like something that should be possible - generate the mesh, analyse it for potential print issues, automatically correct those issues by adjusting the mesh accordingly.

It's a terrible direction for the industry to go, for all sorts of reasons... but does seem like it's going to happen sooner or later.

Wake me up when any firm manages to come up with even a basic autosupport feature that isn't arse for anything above space marine head component size. Because as things stand, just about every single one of those tools you're talking about struggles to deal with even the most basic and obvious flaws. And there are several issues for which there is -no- tool of any kind available and no prospect of one beyond wishful thinking.

If they can't even get presupports right after struggling with it for over half a decade now, I don't see a solution arising in the next ten years. Maybe after that, but frankly, I suspect we'll be using some of those liquid gel/light 3d printers or the like by then.


Automatically Appended Next Post:

 ced1106 wrote:
> unprintable

Good point. IIRC, In early digital, you had these nice looking digital sculpts with tiny spikes and other unprintable bits, even though they were designed by a "sculptor". Eventually, we got to the point where digital sculptors would make miniatures that could be manufactured.

Aye. My partner's a 3d sculptor. So many 3d artists even today have NO idea how to actually translate a sculpt into something you can feasibly print - let alone something that you can bounce around on the tabletop without breaking up. You have

1) 3d artists who never saw a 3d printer in their life and sculpt all sorts of fantastical detail invisible to the naked eye when printed, or too fine to print, or other such issues.
2) 3D Artists who have mastered the above, and make a technically functional file - but which is badly proportioned for the tabletop. Or has detail hanging off at opposite angles to how the print layers upwards. Or multiple thin sections running parallel which will be a nightmare to support. AKA, it works, but not well.
3) Artists who actually test print their stuff, and learn how to sculpt for producing on a SLA 3d printer. That is to say, they sculpt in a manner complimentary to the means of production and purpose. This is no different to physical sculptors learning to angle their posing for two-part mold production and so on. It's the difference between an amateur and a professional.

The AI stuff is going to function like Step 1 above (at absolute best) - because it can't think and adapt how it does things. It's extrapolating off a picture. If a monster sculpt has gribbly stalk eyes that would be too thin to print, an artist can adjust the whole design to account for that and keep things not looking stupid. Even if you program the AI to thicken bits that are too thin, it can't adjust the rest of the model/design to match, because it doesn't understand what it's doing beyond 'make thin bit thick'. A 2D program doesn't -see- in the same way a 3D person does.

And that's for a single part models! Imagine taking a multi-part vehicle kit - something designed to slot together. How is the AI working off a 2d image going to do that? Maybe you can make it slice at the narrowest parts - but are those always the best or most appropriate places for a model to be separated for multi-part assembly? Of course not - it depends upon a host of factors - which an AI cannot adequately understand.

At this point in time no one really knows how far AI is going to go.

Its not knowledge search or functionality that fuelled the AI creation and development... its pure corporate greed and unregulated theft.

Nothing new there, but if there is the slight chance AI will gain relevance in any industry then Devs will train it.
What im saying is printing Objects of any sort is a new industry reaching people homes and will become more and more mainstream so yes "gaming miniatures industry" will be part of that and as such AI will be trained to develop more and better printable 3d assets.

Its not 10 years or 5 years... the speed AI is retraining itself is beyond normal human timings and in around these 2 years you have seen its first gens ugly face obliterating some industries* already.
Digital sculpting for instance did not take 10 years to relegate traditional sculptors to a modest existence.

To finish off how I started, no one really knows how far its going but just make sure you dont underestimate it.

* I work in an Industry that is facing total annihilation even this week we had major portfolio companies presentations coming in and they were genuinely scared of what its already doing to them.
As a side hobby I also sculpt traditionally for wargaming companies for 30 years now and find Digital an extension of my 9 to 5 which is kind of boring so I dont and stick to traditional... and now even that is compromised with AI first gens.
Go to Etsy now and I can find my Original Traditional work, that I sculpted for a miniatures company and they produced it, stolen and shamelessly being sold as STLs... thats reality now!