The method worked well enough, I'd say. Nothing like the image from above but I'm only dealing with 55k cells here. I made a bash script to automatically generate the "shells" with surfaceTransformPoints. I need to fine tune it more so that the flow regions I am interested are well resolved. It has a few artifacts but it's probably because of the coarseness of the mesh. I hope that won't affect results much. Ideally, these shells would be concave hulls so they espouse the geometry well enough but are devoid of "holes" or irrelevant detail. But I also made it very coarse so that I can quickly iterate it before I commit to a higher resolution. I used level 5 distance to refine around the main geometry and then a big stretched STLs as level 4 just to see the outer shell clearly. I would be interested to see if this method can be combined with distance to absorb some of the artifacts This method is quite effective. Thank you for your help. https://cdn.discordapp.com/attachments/397874745464586242/1467243875264495860/image.png?ex=697facd7&is=697e5b57&hm=7afb9233cd8141d08deb18f026727e2c7125f716320369d00d2bc8eb04c5df58&