Awhile ago, I was bemoaning tools for level editing, and said "gosh, I should put my money where my mouth is."
Also awhile ago, somebody - probably Joe Wintergreen - was bemoaning the lack of the ability in a level editor to just grab a door and... move it in the wall, instead of having to deal with nineteen million different operations to do it because you have to take the entire wall apart. This operation - where you have geometry and you want to change the relative scales, proportions or locations of elements - is called reshaping. Anyhow, I'm pleased to put my money where my mouth is:
Our slippage-aware reshaping formulation allows you to select handles on objects and move them to new locations, and also to select points that should stay still; we perform the reshaping indicated by these gestural points in such a way that the style, proportionality, and global features all remain where you want them to be, but such that the things you do want to move or rescale end up moving or re-scaling. This is unlike classical deformation mappers (ARAP and friends) that are primarily designed for posing operations where you want triangles to rotate a lot and scale as little as possible (hence the "As-rigid-as-possible" part of ARAP deformation).
This is a joint research project with the folks at the University of British Columbia who I usually work with; the first author is Chrystiano Araujo, who did the bulk of the programming work. On this one, I mainly did math, writing, and Senior Author Things (mainly looking at things and going "hmm, that looks wrong and this is maybe why?") Paper and code will be up some time in the near future, but for now take a look at the video. I really do think that this is a tool that will make a lot of people's lives easier. As an added bonus, it's ML and AI-free. Interpretable results and deterministic algorithms FTW.
(Incidentally, we had a 2D version of this last year: https://www.cs.ubc.ca/labs/imager/tr/2022/ALUP/ which might be relevant to a lot of game development tasks too!)
