/3/ - 3DCG

>>951384
Animate it without the physics and compare it to real life.

You need to be able to able to rotate the legs 130 degrees forward and about 20 backwards to achieve any interesting pose.

I use this structure with Automatic Weights.

Other than that the Cloth sim and the overall shape and consistency of the object are pleasant.

>>951398
very similar to my setup, I dont like the front deformation I got, but can fix it with some weight painting

>>951400
thanks! cloth sim with tight clothing is kind of a nightmare to work with in Bender.

Everything itt looks horrible. IF and only IF you want good looking sims use Houdini

>>951396
this is the best rotation I can get before it starts to look funky

>>951411
To get past 90 degrees you need to cheat because the Armature system will always lose volume. I put a Force Field inside the Cloth sim to maintain it. If you look closely at >>951374 you can maybe notice the outline.

>>951413
What about changing the scale of the bones during the animation do get better volume deformationg? If we have to cheat anyway

>>951420
you're not scaling an actual bone, you are scaling a curve. If you want this to look right, use houdini. Again.

>>951420
I've tried scaling and I don't think it's useful. To move the leg bone a little is sometimes useful but personally I don't want to deal with drivers and constraints.

>>951384
What's your setup?

>>951529

>>951422
How long before softwares like Houdini become obsolete by AI?
Who cares about real time simulation when eventually you'll just render a wooden stiff animation, and the AI will add in the softness and jiggle later.

>>951559
I thought the fleshy part was a simulation as well because mine is. But it's interesting nonetheless.

The fuck is going on here? Wheres the "ngmi's"?
Wheres the usual 3dmax/maya elitist dickriders?
Wheres the contrarian asshats that bring others down?

This isn't a real /3/ thread... am i in the wrong board..?

>>951565
I thought ngmi was strictly a /ic/ thing?

>>951569
You must be new.

>>951565
The earth is healing.

>>951559
>>951563
Can I see the parenting and stuff?

>>951565
/3/ is fucking dead beyond one or two autistics posting constantly.

>>951637
But I don't think I want to adopt you.
>>951668
It's not dead. It's healing. The Internet is healing.

>>951384
Learn some anatomy. Raise that pelvic pivot upwards, and apply shapekeys to it for the love of god, also why are you simulating tight clothes?? Drive it with a deformer and apply once again corrective shape keys.

>>951711
Can you follow your own advice and show the result instead of just talking about it?

>>951719
>can you do this for free on free time?
No, I'm not a fucking janny.

>>951733
So you're not going to follow your own advice and you're not going to post anything.

>>951739
you realize his experience is probably 1 hour checking out blender and giving up right?

>>951670
BONE parenting.

>>951745
He's a Maya user. I know because I was able to convince him to post something once. It was some seriously bizarre stuff. He's been doing the same shtick for months.

>>951767
Nigger what the fuck are you talking about?

>>951384
it doesnt look like the butt is responding to tension or compression at all. multuple bones is the way to go. It should be the foundation. On top of that you add soft body dynamics.

>>951733
Thankfully this Anon gets more anatomy than you OP
>>952316
my dude that's literally a pelvis, what a nice ass

My latest project features a MC with an outrageous donk.
I'm leaning towards baking wobble into the animations and not running RT dynamics at all.

Its not normal looking. The ass cheeks themselves need to be compressed upwards when legs are moved backwards rather than pinched at the bottom as they are

>>952772
This guy knows asses.

Another attempt, I'm working with soft body simulation this time

>>952829
It's very hypnotic.
Try to do the same thing but with the Cloth sim instead. They're similar but the Cloth sim has more options. Especially Pressure and Internal Springs.

This is where I'm at.

The colliders inside.

Thank you for showing your progress

Working on the front. This is the general idea but I should be able to improve it shortly.
These tests with internal colliders are all Softbody sims because the Cloth sim doesn't work almost at all. The Blender physics stuff is in serious need of bug fixing.

>>953166
Thank you but this is /3/ You're supposed to say: "it looks like shit" or "git gud" or "apply shape keys" or just NGMI.

Small update. Still missing some important details but for some situations it could be good enough.

And from behind. Next I'll try with some Mocap data.

This is not what I wanted. The problem is that the simulation is really pushing my computer and it glitches at every Mocap clip transition.

This is roughly the setup I'm using to stitch Mixamo Mocap animations together.

There are still glitches and it's going to be a lot of work to get rid of those but in the meantime you can have a glimpse at the future of realism: full body giggle.

>>952829
I have been trying to get what you had here as best as possible with using cloth sim/surface deform. Now that I am looking into it more heavily I think I found a video that might help you a little if you continue without using softbody. https://twitter.com/lewdineer/status/1677391911684808727

I tried to make a proxy mesh that would cover the area you had in that animation. Essentially from naval to knees. That will not work apparently since it says that vertices are overlapping or something when I tried to bind it.

>>953963
I didn't properly explain what I wanted to achieve. It's a rather subtle effect.

>using cloth sim/surface deform
The Cloth sim unfortunately doesn't work because of how collisions are implemented there. I've had to read the source code to find out the details. The manual isn't helpful, but long story short, the Softbody sim is the only one that can work properly with the kind of internal colliders I'm using.

The flickering in the top part is because I haven't made a rib cage, shoulder and elbow support yet. It's fixable.

Other than that, there is gravity and the ground is a collider as well so every time a foot hits the ground it creates a small wave that propagates upwards.

Overall it creates that particular female consistency while walking, which is what I'm after.

The problem I haven't solved yet is that any small inconsistency in the Mocap data upsets the simulation a lot, because the simulation keeps track of the velocity of every vertex. So I'm probably going to have to have to write some special filter.

I'm sorry I don't have a twitter account so I could't see your link.

I use decimated meshes to speed things up and as a result, the collision balls approach that the Cloth sim uses doesn't really work.

The Softbody sim does it the right way except, because it's Blender, the outer guard is bugged. There's actually a piece of code missing. Pure Blender programming quality. But it can be just set to zero.

I really want the simulation to be active on the whole body because I think it's going to help me with the details in certain situations.

>>953722
It is well within acceptable parameters. Up the dampening to smooth out the vibrating. It's looking good enough to do what it needs to do.

>>954035
The dampening was already set to 1.
The glitches are mostly due to imperfect joins between Mocap clips.
The flickering is mostly due to the maximum number of iterations per frame.

Next time I'll choose a Mocap with more action. We'll see what happens. And the fact that the whole model is a simulation means that it should react to external colliders as well.

>>954055
Damn, youre actually slowly getting there, this looks workable

Just out of curiosity, this is from Blender 2.37a. All the simulation features I've used for these little animations are there.
blender-2.37a.tar.gz 16-Jun-2005 8MB

Samba dancing

Skirt with the Cloth simulator

A couple of notes if want to replicate these experiments on your own.

Decimate your model to the lowest acceptable resolution before running a simulation on it. If you leave it as quads it's going to be way too slow. In my experience, 50% ratio is always fine and while you're at it, you can also triangulate it, so you don't have to use the Stiff quads option later, which is expensive.

Both the Softbody and Cloth sim have a special case where if you set the goal strength for a vertex to exactly 1, that vertex will be excluded from the simulation. You can mask off the hands and the head that way and it'll give you better performance.

The next experiments I'm going to make I'm not going to be showing them here, but I hope somebody picks this up and make some nice things.

Somehow I was able to re-target this:
https://github.com/BandaiNamcoResearchInc/Bandai-Namco-Research-Motiondataset
It's really high quality. I especially like the anime girl walk at the end.

>>954155
with the clothes on it looks horrible...

>>955054
It's a woman, of course it looks horrible with clothes on.

hmm anime

Experimenting with Force fields.

>>955068
Japanese once again proving that simple, artful approaches are always more appealing than highly technical, tasteless ones.

>>955554
Ok, so you can post some of your finely crafted creations or you can kindly fuck off.

>>955554
IF your character has a nOrMaL butt. Making a flat ass is easy because whether you are using vertex driven deformations, twist joints, or dynamics the magnitude of any is going to be very low and any instability will be hard to notice unless the camera is right up the character's ass.
Real pioneers in 3D are trying to solve the big boy problems of absolutely massive asses.

To find the parameters to make this work was really hard and it may not look like much but there's a lot going on. You'll see.

>>955574
that looks stupid as fuck its not even hot

I'm trying this.
There are no support loops so far, basically just an extruded cylinder. It won't deform nicely without blend shapes, will it?

>>955859
> No camel toe
Duuude

>>955759
Model looks like a man with gyno and no ass. Do better?

>>955861
Fun only after work, anon
Topology must be correct to model the details later!

>>955863
I don't think you like women.

>>955859
Seriously, what can be done better?

>>955941
A rig and an animation?

>>955859
The edge flow is 45 degrees off to the angle it's going to be deformed the most. Making your job more difficult. The edge loops should unfurl like a paper fan. The edge flow looks very nice and smooth but frankly this part of the human body doesnt flow nicely. Muscles, bones, and tendons are all criss crossing and making abrupt changes in direction
>>955788
you look stupid and you're not even hot

>>955759
that's the battle you chose when you chose dynamics. For the rest of time you'll be tweaking the same parameters back and forth trying to solve the same problems over and over.
It's seductive. "hey let the computer do the hard work" but then you're stuck with an infinite amount of "easy work". At some point you have to decide it's "good enough" and do it before you've sunk more hours getting it that close that you could have just made blend shapes and twist bones.

>>955951
And can you show me some examples of your "blend shapes and twist bones" for comparison?

>>955759
This is horrible

>>955957
There are no corrective blendshapes here. Just a single twist bone. It needs it because the volume loss is proportional. Bigger butt more apparent shrinkage. My project uses model swapping for changing outfits. So I either need a solution that's 100% joints or I have to make blend shapes for every clothing mesh.
For your butt your problems are very simple. I bet 3 corrective blend shapes would both solve 99% of your deformation problems.
>lifting the knee all the way
>leaning knee back as far as you're planning on going.
>leg all the way out to the side like a french split.
Then you can apply a very light dynamics systems for subtle jiggling rather than controlling the whole deformation of the butt. Then fully dynamic clothing on top.

>>955976
here's me setting up a throw away corrective blend shape. Just to prove I can do that too in case it's necessary.

>>955976
>>955977
It's the same stuff you've already posted 6 months ago. Honestly, I can't even tell you if it looks good or not because it's not a human shape and therefore I have nothing to compare it to.

>>955977
This is not even close to industry standards

>>955977
Sculpt the posed shape and extract the delta. No need to trial and error it.

thanks for showing your work

use x muscle system, saves a ton of work if you want to have aesthetically pleasing movements.

>>955988
You wont have sliding muscles underneath the skin. It will look horrible and not industry standard. Don't be lazy

If you want to experiment with you own rigged models with the Softbody sim, these are roughly the parameters I'm using.
In short, keep #1 a low as possible and use #2 as your main source of friction.
Also, if possible use the Cloth sim instead because the Softbody sim you really need to read the source code to be able to use it.

>>956022
You dont want to use either of those. Use industry standard FEM.

>>956023
I don't use Blender because it's good quality software. I use it because I'm a programmer and it's open source.
We're talking about two different things. You're talking about quality art, I'm talking about physics experiments and research.

>>956028
You want to use a FEM solver, not a throwaway blendshape on skincluster. What the hell are you even doing

>>955979
Yeah i keep having to make the same point and prove the same idea over and over. Why do i need to make a new gif everytime?

>>956019
No you wont. That's not what corrective blends do. Youd need a muscle rig for that.

>>956047
>i keep having to make the same point
Who's making you do that? The voices?

>>956050
Wow youre really upset that i spent my own time 3d modelling a big butt.

further research

>>956055
I'm not upset. I just think you're mentally ill.

>>956060
Youre seething because i placed vertices at coordinates you deem improper, and youll never ever post your work either.

>>956065
I don't blame you because you're mentally ill. I blame the jannies because they've let you harass people who post their work around here for at least 6 months.

>>956071
You are maximally ass mad by maximum asses. Youve created a false persona whose "too good" to give real advice or even post your work here and your careful delusions are shattered by anyone whose trying, even a little bit.
Look at you. You cant let it go that someone made a cartoon you dont like while also being 100% correct on the topic at hand.

Hey OP, sorry to shit up your thread but let this guy serve as a cautionary tale. If you try and solve all of your problems with dynamics youll go crazy and never finish any project. Never have anything to post. Be forever angry.
Twist bones and blend shapes. There is no counter argument just kvetching.

>>955950
What is the best topology for this type of mesh?
Every mesh seems different. Where can I learn about correct topology for characters?

All places I've read just mention that I should use quads as often as possible and try as much as I can to make limbs as "cylinders" - but then I lose the landmarks.

How can i find the correct dosage of quands, correct topology and detail?

>>956089
Pic rel

>>956089
>>956090

pic rel 2

>>951561
Who cares how long, I can garuntee you're ngmi even when those tools become public. You cant even be fucked to learn something new now, you wont later either.

>>951565
>worry's about other software
>blender user
>cums during blender faggots stream

ngmi

>>956057
disgusting

>>952751
Holy smokes!

>>956091
I think you have faces too densely packed where there isnt really any relevant model detail or deformation happening.
Try to normalize the resolution a bit more so the butt gets more and that lower back gets a bit less.

Get your girlfriend to let you retopologize her waist-area with a sharpie to learn how to 3d model this area.