I was curious about the use of AI for texturing. For example using it for making simple building assets just by extruding the faces, etc, or any other cases. Do you guys use it, why or why not?
Nope! In my opinion, generative AI has absolutely no place in the creative workflow, especially by artists and fellow creatives whose artwork it has stolen, and whose livelihoods it puts at risk.
Just my opinion, though. ![]()
I would agree with @Magnavis, AI has no use in Blender, apart from potentially giving instructions on how to diagnose or fix something.
I am not going to write the same thing over and over again, so I have put some links below to what I have written before:
It’s hard enough “to read” some topology from real images and if you have ever used multiple reference photos of the same object you will see: they often do no match, simply because they aren’t orthogonal..
So when using something hallucinated images.. then it is even harder “to read” the topology, because they do show “different hallucinations”.
In the end you have to work more on something to fix it than doing it right in the first place.. this also applies to experienced users. Somewhere in the workflow using real photos you may think: “Oh what is that ?” but because it is “real” a human can “imagine” what this is.. especially with experience. But windows over corners in different highs or chimney where the are useless or some other hallucinated weird thing.. just gets you out of the flow the build those houses you mentioned.
So there is a huge difference between some creative imaginations of a human and some hallucinations from some AI. Even the language we use gives a hint which images are more “real”. ![]()
I hate AI. But, I have used it for making textures. Let’s say I have 2 product images of a carpet:
And I need a high resolution image for a texture because the product photos from the manufacturer are pathetic(not a positive word, but… well it is what it is
). The image generators can take images as reference and generate this:
I can then photoshop a high resolution texture from this bit, since I have a corner and so I can duplicate that 3 more times and there is some area that I can duplicate a few times for the middle of it.
It’s not perfect since I was in a rush and did not bother to fix issues with the seams of image stitching, but that could as well be fixed with more work(the manual kind that involves no AI
).
So… I don’t know… It can be useful if you have specific input. It’s not really about making art, I just have a product I need in my visualization and the manufacturer didn’t bother to share high resolution photos of it.
I don’t want to be making a carpet texture from scratch all day…
The issue here are these uneven dark or light bits, but that’s an issue with my manual work and they are indeed definitely fixable so it could have been a really decent texture. ![]()
Letting the AI alone texture something… No. It cannot. The results are crap and unfixable with any reasonable amount of manual work.
I’m also against AI in visual arts, but I think this is a completely appropriate use-case. If you give it a very constrained task where it’s easy to evaluate the output and adapt it to your needs, it can be really useful.
This is the only usecase I sometimes use GPT (free account). It often saves time and eliminates the need to search for textures. This is all for hobby use, nothing commercial. Otherwise, I don’t use AI in the graphics/3D sector.
Translation:
I need a 2D texture to map onto a flat surface in Blender. It’s a wooden table with signs of wear and tear, a few cup marks (water stains), nicks, and scratches…
It looks great. Is it possible to make it seamless?
Yup completely agree with you too. I also got quite scared because of ChatGPT’s new “Astra” model. It seemed to do everything quite well, although I haven’t seen anyone show the topology of the models that were generated.
I was also wondering if its still worth it in the long run (doing 3D), because of course Astra may have downsides, but AI is improving day by day, and its making me worried if I ever wanted to get a job in the 3D sector, etc.
Yes I also agree with you. In general, if I ever do want to find some images that I want to use for buildings, etc. I usually use Pinterest, although for me sometimes its difficult to find the angle that I want it to be at (mostly flat front view). I heard theres a app, or an addon in blender that makes it possible for you to align a image thats taken from a different angle, and like you know where to extrude, etc. (sorry Im bad at explaining, and I forgot the addon name, also trying to find it)
I was not aware that it does a somewhat passable job at making a texture seamless. Seems to struggle a bit at a full PBR workflow though.
Against my better judgement:
Hope you don’t mind that the scratches and water stains are raised.
looks good ![]()
There was a website that could also generate normalmaps, although I don’t know how good it is at doing it.
Incase you want to try it, https://www.normalmaponline.com/
But these workflow: using soem image and use aome filters to make “fake” roughness, glossy , bump, normal maps.. is ooollldddd..
Also the “make tileable” features in image apps…
So why do we need AI for this ???
Even image resize filter were a thing before AI…
“if I ever wanted to get a job in the 3D sector, etc.”
If you want to get a job, you will need to be able to create all your own textures to show the skills that are needed. There are amazing things available that can help do this, like actually modeling the patterns that are then made to be seamless and tileable for use in your texture, layering different procedural textures to reveal your seamless bits, etc.
AI can’t just do it for you and then you learn how to do it from that - you’d need to have the AI teach you step by step and then you’d be up the creek because it would lie to you about how it got that result. Best bet is to study hand painting materials so that you can appreciate what you are actually seeing in others’ works.
My 3 cents, keep the change.
No, I don’t think we need AI for any of this.
Yep thats what Im planning to do, Im planning to learn substance painter this month.
What I actually meant by saying “if I ever wanted to get a job in the 3D sector, etc.” is if 3D artists are ever gonna get their jobs revoked / replaced. I do Blender as a hobby, and find it really fun, creating enviroments, scenery, etc. With AI improving day by day, will 3D artist be replaced? (companies cutting costs by firing 3D artist since its cheaper that way, and what not)
The difference here is that ArcBrush predictably uses luminance to determine height, while ChatGPT instead takes a wild guess.
Key word
![]()
Found the addon, it was called “fSpy”










