>>2802152>>2802108>>2802032Tested a soft body dress (without a proper rig) and also a coat:
https://video.twimg.com/ext_tw_video/1697714672625917952/pu/vid/1280x720/CsHEKzpRS4E4nOj-.mp4?tag=12Findings:
1) As long as it collides with existing rigids, my system (RTX 2060 6GB, Ryzen 7) can't handle processing the physics in the viewport (but a cloth freely moving through the air works smoothly) but can still render a video (which was new to me), albeit very slowly even with a simplified model. (On a normal model, it took 386 seconds to render 300 frames of a simple project with only the model, motion, and audio loaded, a 3500% increase in rendering time compared to 11 seconds for the same video with physics enabled but no soft bodies (104 rigids, 48 joints, 126k vertices, 174k polygons, 646 bones, 942 morphs, 1 soft body). It's like I had a full RayCast setup with a bunch of lights and at least two models.) Rendering the test video of 5600 frames (3 minutes 6 seconds) took 2 hours 16 minutes 53 seconds, whereas with softbody physics off it took 3 minutes 46 seconds: a 3600% increase in rendering time! I don't know if it'll scale like that with a Ray-MMD set though; there should be a cap to how slow it can go.
2) Needs a high poly mesh to not produce garbage (look very closely at the red dress). Even then, it still doesn't produce results that justify using it for clothes.
3) It doesn't require weighing the whole material to a single bone, like a deviantart tutorial suggests, which is a relief with how it functions. It actually needs a properly rigged unmoving base if it's clothing.
4) Soft body physics are not compatible with SDEF weights. Even if you include the elbow areas of the coat in the fixed vertex array, they will still dislodge and sink.
5) The coat clips through the body with even a collision margin of 1 (the video has 2, and the result is ugly; I will not retry it in the next few days, but maybe I'll retry with the dress).
