🧵 There are papers on COOL REAL TIME PHYSICS AND 3D GRAPHICS dropped literally EVERY FUCKING DAY
Anonymous at Fri, 31 May 2024 17:22:55 UTC No. 985098
...and yet almost nothing of that gets implemented anywhere. Especially, Blender:
>I sleep (for years)
WHY
Anonymous at Fri, 31 May 2024 17:48:13 UTC No. 985100
>>985098
Those who really wanna use it read the paper and roll their own version.
Anonymous at Fri, 31 May 2024 17:49:52 UTC No. 985101
>>985098
It's the same with Maya. nDynamics is ancient at this point. And now they're working on Bifrost, which is utterly useless due to how heavy it is, unless you can afford a rig that costs $20k. I really should just learn Houdini already, for my simulation needs.
Anonymous at Fri, 31 May 2024 18:07:43 UTC No. 985102
>>985098
>WHY
Blender's development depends on what the Blender Studio needs for the shitty short movies that they made once in a while, nobody there gives a fuck about indie game development or whatever the fuck you are playing with today, that's why Cris.
Anonymous at Fri, 31 May 2024 21:02:44 UTC No. 985107
It's because failed "artists" get extremely upset at physics people because they *demand* to have jobs as low quality human physics simulators If nothing else, AI will fix this issue shortly and permanently. Jannies are doomed either way.
Anonymous at Sat, 1 Jun 2024 04:39:55 UTC No. 985139
the thing that infuriates me the most that apparently most 3D software has physics that try to be as accurate as possible while being slow as shit.
meanwhile games have physics that tun at 100fps that are rough, APPROXIMATE physics but in most cases that would be PERFECTLY fine for animations. you don't need autistically accurate physics every time that take ages to compute. fucking give us the game physics.
Anonymous at Sat, 1 Jun 2024 05:57:00 UTC No. 985143
>>985098
industry software is stuck with decades old shit because the studios have custom scripts they don't want to rewrite every few months when some new tech appears. autodesk has been adding some AI functions lately but it's still mostly gimmick no one serious would use. blender is irrelevant garbage for retards to waste time in.
Anonymous at Sat, 1 Jun 2024 06:04:33 UTC No. 985144
>>985139
Well, then if what you say it's true instead of being infuriated all the time you could start here:
https://github.com/jrouwe/JoltPhysi
or here:
https://github.com/bulletphysics/bu
And make your own dream physics simulation environment.
Anonymous at Sat, 1 Jun 2024 19:30:25 UTC No. 985195
>>985098
Test
Anonymous at Sat, 1 Jun 2024 19:59:52 UTC No. 985198
>>985101
bifrost is not useless, you just don't know what you are doing
Anonymous at Sat, 1 Jun 2024 22:10:30 UTC No. 985212
>>985098
Why don't you make your own?
I'm working on mine at the moment
Anonymous at Sat, 1 Jun 2024 23:21:43 UTC No. 985229
>>985198
No one knows how to use it, YouTube videos show unnecessary steps in creating it. The only real way is by replacing the model. Anything beyond the browser is junk. Autodesk has assumed we know every command code to build any SFX.
Anonymous at Sun, 2 Jun 2024 01:15:27 UTC No. 985232
>>985144
Doesn't Blender have bullet physics that are somehow completely crippled and slow?
Anonymous at Sun, 2 Jun 2024 01:44:21 UTC No. 985234
Before I forget, since for whatever reason everybody wants to do Softbody physics with nodes these days:
http://nortikin.github.io/sverchok/
I've tried it a while ago and it kinda works. It also has a non-time-dependent mode to do simulations for tensile structures. If you can stand Sverchok, it's also available for Blender 2.79
>>985232
It's there including the real-time softbody parts and it's not slow or crippled but it's in the game engine. These days you have to get it from the correct version of Blender which is:
https://rangeengine.tech/
Because the Blender developers removed it from the mainstream distribution. They did that for reasons not yet known to humans.
Anonymous at Sun, 2 Jun 2024 02:33:54 UTC No. 985236
>>985234
Pic related in case you don't believe me.
So, at some point, the wise Blender Developers had a choice: do we port that functionality which already exists in the game engine over to rendered animation part of Blender or do we rip everything out and replace it with something that doesn't really work, never really worked and will probably never really work because, fundamentally, we really do hate our users?
Anonymous at Sun, 2 Jun 2024 21:27:07 UTC No. 985292
>>985198
I said it was useless, unless you have a high end rig. I admit i exaggerated a bit, it's not 100% useless with a lesser rig, but it sure is not a great experience to use it on anything that is not a threadripper.
I know how to use it, I've used it several times. I'm not maybe an expert in it, but I can get okay results out of it. My issue isn't with its usage, I actually really love the node based graph for it. What I don't like, is that anything with sufficient quality, generates at the speed of 1 frame per minute, unless you have a high end threadripper with a minimum amount of 64 cores. You can create with lesser hardware as well, but it's frustrating and incredibly time consuming to leave a simulation calculating for the night, only to find out it looks like shit the next morning. You can get close(ish) results with using lower voxel resolutions for initial testing, but you still need to test out the final results, before you commit to rendering, which takes forever, unless you settle with something lower in quality.
Anonymous at Sun, 2 Jun 2024 23:03:48 UTC No. 985303
Jesus just use Houdini if you need a decent bullet implementation. It's incredibly fast, especially if you take the time to optimise your SIM geo.
Hell, you can do very fast rigid body Sims with the xpbd solver (vellum) these days as well.
The new hotness is definitely that vbd paper and implementation that came out a couple weeks ago. Definitely need that shit in Houdini asap. It's so fast and robust on soft bodies.
Anonymous at Sun, 2 Jun 2024 23:45:07 UTC No. 985304
>>985098
Why don'tcha link us to some of your favorites? I've been excited to finally have a coding tutor (AI, lol) so I can experiment with gfx. Bet a lot of people are. It'd be nice for a QRD of developments over the past decade or so - I grew up with Bryce and Poser, graduated to Maya, then had to sideline it all for my career. But I'm a NEET now so I've got time for fun stuff.
Anonymous at Mon, 3 Jun 2024 00:27:30 UTC No. 985306
>>985098
where can I read them?
Anonymous at Wed, 5 Jun 2024 10:36:10 UTC No. 985550
when i read papers and then give away implementations of the advancements for free everyone shits on me and says 'the big studios are not doing it for a reason therefore you are INCORRECT'
Anonymous at Wed, 5 Jun 2024 11:00:08 UTC No. 985554
>>985550
What "implementations of advancements" have you given away for free so far? Can I see one of them?
Anonymous at Wed, 12 Jun 2024 23:56:27 UTC No. 986592
>>985550
wat
Anonymous at Thu, 13 Jun 2024 16:21:15 UTC No. 986658
>>985554
He means that he's screaming at developers to do something on a paper that he doesn't understand.
Anonymous at Mon, 15 Jul 2024 17:55:58 UTC No. 989868
>>985098
sure
Anonymous at Wed, 7 Aug 2024 00:11:51 UTC No. 991711
>>985195
failed