Post by Thrace Kelsick

Funny that in the age of automation and AI, we still don't have a solution for retopology that holds up to scrutiny. Not a big concern for common repeating organic shapes such as humanoid, feline, canine, etc. as re-using base meshes works most of the time. But hard surface remains a unique challenge, especially for translating CAD into subD friendly polygons. So, over the last few days I decided to take a deep dive into what's available in 2026...and found many of the same old shortcomings😅 1. Poor circle/cutout handling 2. Innappropriate mesh density (dense/sparse in wrong places) 3. Odd number of spans/edgeloops, when 99% we want even numbers 4. Spiraling edgeloops (bigger issue than they might seem) Even with an inordinate amount of input (user defined polygroups/smoothing groups/hard edges, weight painting, etc.) and symmetry, the results are underwhelming across the board. I tested AI tools like Tripo 3D, standard DCCs like Zbrush, 3DS Max retopology toolset, and Quad Remesher plugin for Maya. Let's take something simple and everyday hard surface as a bolt, we still run into quite a few problems! (see below) Can auto retopo at least be used as a base? Yes, but in the end I find there's usually regret: Spirals not discovered until UV layout. Odd count of spans in circles. And before you know it, the "touch-up" becomes a laborious process that, along with setting controls at the beginning, could equal a fully manual timeline🤷‍♂️ #retopology

Video Content