Do you all noticed that "Human Generator" created meshes can be erm...identified?

General Forums Blender and CG Discussions
1

Lately due to the brief appearances of model importers having their imported models featured in the forum, I noticed something that seems to still hold true even now.

Even back then during the early days when there was only poser, you could tell that a model was from poser, as more programs appear ~ iclone models are also obvious, and recently a featured daz model was called out subtly right off the bat by fellow forum members and fortunately, the model importer user didn’t deny it.

Make Human Models also have that erm…Make Human look…

Does any of you notice that as well ?
What do you think is the main reason why character generated meshes always externally visually look like they are generated from the program they are generated from despite all the parameters they can tweak ? (Rather oxymoron a statement I know but you get the point.)

What do you think are these program’s greatest weakness ?

2

Yeah. It’s not limited to characters really, anything that can be “generated” can have that makeshift common look. There’s no real getting around it except by learning how to model characters / objects yourself.

Personally, I find it gets boring having everything look the same and I feel too many people are trying to copy that Pixaresque “haha look at me” kind of look. They also tend to have big goofy eyes, lipless mouths, nostillless noses, big grand smiles everywhere, and if they’re animated, big personalities… No computer is creating those but it would be so easy to create one! I love looking at work done on this forum cos they don’t always follow that formula, even if they’re using tools like Makehuman. They are there to be used at the end of the day (though I do wish more people learned the raw skills to model / animate for themselves.)

3

I mean, when you’re using the same topology, same textures, same shaders… you can tweak that as much as you want, but you paint a duck whatever color you want and it’s still a duck. I personally don’t consider character creator software art- art is personal, it has a soul, mass produced art isn’t art to me.

That sounds elitist, so let me say- you can make art with character creators, but you can make art that’s a lot more meaningful and personal without.

My opinion aside, if people want to use character creators, they should, but I think there’s really something to be gained from doing it yourself. It has a human touch that way you just can’t get with one of these programs

4

I’ve modelled a few people, maybe half a dozen, but what if you need a dozen people? I have noticed low poly photo scanned ones are popular, because they have that ‘real people’ look.

5

Depends on your use case. If you are doing interior architecture and you want some people for scale, low poly photoscanned models are great. Often, I will use photoscanned models, but I will remove the textures and render them as transparent grey silhouettes.

If you are doing character art and the person is the focus of the image, it’s more important for your work to be more central to the image.

2 Likes
6

I hope at least the staff here at BA are not going to impose a rigid view on what art is for the artwork section.

I'm sorry, but I will have to lock this thread due to the art being no more than eye candy, please add some elements to the scene to make it tell a story so the thread can resume.

I recall that during the heyday of CGSociety, the staff there actually had the power to declare if a work can be called finished before the submission went through (even if the artist put a lot of work into it already). I would not want to see the same type of thing from another angle here.

1 Like
7

That was from CGSociety?

8

The vast majority of what I say here isn’t an official staff position. Official posts made by me or anyone else on the staff have an orange background or are clearly specified as such. Anything else is my own personal opinions and thoughts. If I’m speaking as a staff member, I make a concerted and deliberate effort to make that clear, and I don’t see anything in what I said indicating that was an official statement.

The staff doesn’t even decide what gets featured, that’s entirely community based, no need to worry about an authoritarian regime here :sweat_smile:

9

I never said that was going to be the case, just saying that I hope that no decision gets made in the future that it will become the case (in response to a hypothetical explosion in generator-assisted work amid the AI revolution and the vast expansion of Blender’s procedural tools).

You seem to be sure it won’t become the case, so I will take your word for it.

10

IMO we need to start issuing demerits for bevel resolution violations.

1 Like
11

@joseph, that’s why I did not use the word “artist” when I describe them, I am being generous by describing those small particular “featured” image makers as “Model Importers”.
The word “artist” never came to my mind and I will never use it on them.

What ? You imported a bunch of assets, arrange them together, shine a light and use the realism of Cycles to render what did you say ? Architecture “art” ?
Yeah Boolean boxes are so tough bro, somebody give them the stunning and brave award.
What ? Your angular box have a curve at the side and you imported some potted plants and place them next to your imported bed ? Somebody call the noble price for the greatest artist that have ever lived and let’s get him featured !
The best I can call them are model importers, don’t expect more, but then again, what do I know, three blank canvases are sold for millions:

Just don’t expect me to call these model importers “artist”, I don’t have it in me to do it…I can’t…
Sure for production work you DO THAT to get your work done ON TIME, but for portfolio/passion work ? Please…

Anyway back to the subject ;p ;p ;p
@joseph You have a point about the same topology/texture and shader.
So maybe a fake artist genera~, sorry I mean a character generator that offers these abilities might be able to make it look like it’s original work.

Maybe I should start looking into these programs seriously…wonder which one offers the ability to change the basemesh texture/shader/topology of the mesh within the program itself.

Confession time: I must admit I am a hypocrite here, I study topology…a lot, and I play with different topologies to see them in action, recently I finished sculpting two humanoid meshes but during manual retopology, I copied/matches 1:1 the topology from topologies I have seen in pinterest, youtube and even models I have bought to study, so in a way, I am a fake too, even though I created it myself, the matching topology makes me feel…like a fraud, like I am “dirty” in some way, and I also used ear or hand parts I have bought and integrate it into my model, so all in all, I am also a fraud, not all my models are “pure” per say.

12

Going back to the original topic, I think the reason most of them can be identified is because most of the time, people don’t make any changes, and simply render/use the output result.

MakeHuman for example is super easy to identify by the textures and default hairstyles to anyone that’s even a little bit familiar with the program. You mention Daz - which can be heavily customised - but again, most people don’t bother to change textures, and so stylistically, you can always spot a Daz character.

On the little tangent:

There was a featured post a little while back where someone literally just exported a generated character and pressed render. People were able to identify it by the skin texturing - specifically the placement of some freckles. This then forced the OP to go and make “changes” which made the piece look worse IMO. The frustrating thing there was that if you clicked on their profile, you could see from their previous post where the artists actual level was, and it was nowhere near the level of this exported character.

Finally…

IMO, there is nothing wrong with this at-all. Topology is…well, just topology. You still need the underlying mesh, which you’ve more than likely created. Almost all facial topology guides are more or less identical, because there is an optimal way of doing it, and people just follow that method. Copying topology isn’t cheating, and certainly doesn’t make you a ‘fraud’. Heck, most professional artists will simply re-use a base mesh to either sculpt directly, or to quickly retopo a sculpt using a program like Wrap. It’s efficient. Why re-invent the wheel…

3 Likes
13

Copying topology is just like copying lines of pencil on a paper. It’s simply the foundation of whatever it is that you’re modelling. Experienced people start drawing characters in practically the same way now (with some differences like exagerated shapes) before it diverges onto other paths like anime, realism, cartoon, etc.

I do find it interesting how easy it is to make a scene with other peoples models and materials and create a unique image from them. It’s not like painting in that sense whereby you don’t need to create absolutely everything yourself. I suppose the hard part in that sense is finding matching pieces. Personally? I prefer to try to keep other peoples work at a minimum and develop my own. No licensing problems that way. Textures are usually an exception though.

I really don’t like using MakeHuman cos I feel like it takes too much control away from me (depsite all the sliders and options.) The best result would be to just start from the ground-up yourself. It’s a great short cut if that’s all you want, but you’re not learning how to create models by using them and you’re doing both yourself and the viewers of your work a bit of a dis-service (cos it’s normally so obvious what you’ve done.)

2 Likes
14

it depends on how the the prefabs are being used.
I use both Iclone/CC3 and Daz
I even have began using Sketch fab models rigged with Auto rig pro

However I am a Character animator.
I rarely post still image “gallery art”

Do I possess modeling skills??

Well… I modeled everything in this clip EXCEPT the Iclone base males hidden underneath my custom armor
and two of the guard sare legacy, low poly Daz figures imported as simple FK rigs with mocap

Now in this 8 minute Blender short, I again used a combination of Iclone& Daz figures and environments

Can tell which is which??

Does it matter to my story Narrative?

Finally this clip contains alot of my original modeling for the environment along with one Character model from sketchfab rigged by me with ARP and some Iclone people imported pre-animated from Iclone.

again Does it matter to the story narrative?

15

Using pre-fabbed stuff allows us to do things on a scale that would otherwise be an impractical investment in time. Not to mention, not everyone has the knack for organic modelling. I used to be a purist, but I have loosened up quite a bit.

5 Likes
16

For Make Human, even I have spotted a few cases where I could see it was being used. Now there are other generators, be it “Human Generator V3” on the Blender Market, or MetaHuman for Unreal Engine. Can those models also be relatively easily spotted?

1 Like
17

Like already said in this thread, it’s pretty easy to copy topology then just project easily available sculpts onto it. You can even use multiple sculpts and then just shape-key between them. Re-texturing might be a bit more tricky.

My personal pet peeve is that all Metahuman eyes look exactly the same and nobody else seems to be bothered by this.

2 Likes
18

I understand that it is easy, my question was whether it is necessary for more advanced generators.

That’s good to know :slight_smile: So retexturing a very little part would solve it.
To me MakeHuman characters always felt kind of stiff, but I am not exactly sure why. I don’t have that kind of impression with other generators though, that’s why it isn’t that obvious to spot the.