Anonymous at Fri, 21 Oct 2022 06:44:32 UTC No. 922700

Came back to it after a break and found that you can merge vertices at certain points and just remove some of the edges flowing into the merged point, and that's an easy way of lowering resolution via a diamond. It's easier to start at full res and reduce from there, it seems that in general it's more intuitive to have a lot of vertices to merge together, than to try to extrude individual edges and then fit them together like a puzzle. In any case I feel like this method works with an arbitrary difference in number of edges between the two loops you're trying to connect, but I'll need to see if it works in a practical case like connecting body parts together

Anonymous at Fri, 21 Oct 2022 14:00:11 UTC No. 922731

Google "edge loop reduction flow".

Also it may be better to learn as you go in practice.

Anonymous at Fri, 21 Oct 2022 17:41:46 UTC No. 922748

>>922680
All quads is not always necessary if the object isn't deforming.

Anonymous at Sun, 23 Oct 2022 09:16:48 UTC No. 922940

https://youtu.be/rwW6HpOcAuw
this vid or J Hills vid on retopology prob explains most of it. other than that just follow gamdevs or ppo on twitter who talk about tricks they do at job sometimes

Anonymous at Mon, 24 Oct 2022 00:06:17 UTC No. 923016

Practiced bridging cylinders together, really messy but now I mostly get how to connect them regardless of the number of faces.
I figured this would be an optimal way of modeling characters (first making loops to establish where different parts ought to begin or meet with one another, and then filling in the blanks, like in this pic: https://topologyguides.com/assets/img/hand_loops.png ), but it's really slow compared to just extruding edges and fixing the topo up later. I guess geometry and the actual shapes/design should precede correct topology, you should have as many faces and edges as you need to support the details you want on the model. Without having the design figured out you wouldn't know how many faces there would be in a given loop or how the loop should flow, so these exercises I did with cylinders would be a horrible way of really modeling something

>>922731
>>922940
Very useful, thanks

Anonymous at Mon, 24 Oct 2022 00:08:37 UTC No. 923017

>>923016
It's good knowledge to have but ultimately, you would do this in retopology. When people add boleans and sculpt them into characters those are used for the high res models.

Anonymous at Mon, 24 Oct 2022 00:27:44 UTC No. 923018

Also wanted to ask if there were any more fundamental ways of reducing faces (while maintaining quads) than pic related. I think logically this would be all you'd ever need to direct flow and reduce loops, but that 1:2 sometimes gave me problems when I was trying to connect odd to even numbers of edges, since it makes an additional horizontal loop that i wasnt always able to close

>>922748
I'm mainly going to do characters so it's important to me to be able to keep quads as much as possible, although I know there are a few instances where use of tris is actually advantageous for maintaining volume while deforming, like on the sides of a knee or elbow. Otherwise, I guess like 5 star poles, tris ought to be contained far away from any deforms by default

>>923017
I don't like sculpting but it's seeming more and more like I'll need to get over that it if I want to make really high res meshes. Though it's interesting that the general starting process for sculpting and low-poly/boxmodeling is the same, just on a different scale; have to feel out the form either by moving individual verts and edges, or large chunks of the model at a time