3D Level Of Detail (LOD)

garrettia

Hi.What is 3D level of detail (LOD) ? What are differences in using Vfx vs games? I would appreciate if you could suggest a source.Thank you

Kanni3d

https://lmgtfy.app/?q=what+is+level+of+detail+cgi

https://en.wikipedia.org/wiki/Level_of_detail_(computer_graphics)

gnoop

I never worked for vfx but my understanding that in vfx it's mostly just subdiv modelling . more subdivided models at close vs less at distance. And rather proxy shapes in place of hires meshes in a viewport.

In games LODs are critical part of rendering pipeline and main part of optimization for better frame-rate.

Triangulated low poly meshes with long triangles optimized for games are usually not subdiv models. Also baking normal maps the way they would produce similarly looking shading on simplified next lods is a key for game modelling while in Vfx the first next lods might be to far ahead to care about.

Unreal has AI based lod generator but usually AI solutions is not good enough for everything but still good enough for many typical things. For mobile games or sport games requiring good fps or games just not on Unreal engine where squeezing a few extra fps from better geometry optimization is very welcome. Thinking in few LODs levels ahead is an important skill.

Hierarchical LODs or "grouping" as well as Instances grouping are important part of modern game engines too. When several separate objects are grouped into single next LOD objects to decrease a total amount of objects to sort and manage. Sometimes at preloading time , sometimes on the fly . It's something you need to understand and use if doing level design.

An artist concern of all this should be modelling techniques that are not relying on extra mesh tessellation and support loops (which would be gone in next LOD) to produce proper surface shading . I.e. you should be able to produce clean geometry without shading artifacts, gradients . Usually a combination of split/hard edges, face weighted or edited (explisit) vertex normals does it.

The later is my opinion, not rule, based mostly on my former mobile games and ancient Voodoo cards modelling experience where evry vertex was at count . Now my role is more of material artist than modeller.

Thinking LODs ahead and using vertex normals tricks certainly takes extra time so in modern AAA game pipelines you may find support loops much more acceptable and thus a dominant way to fix shading issues. It's a kid of time/cost vs fps / quality issue as basically everything in video-games.

sacboi

"vfx it's mostly just subdiv modelling . more subdivided models at close vs less at distance. And rather proxy shapes in place of hires meshes in a viewport."

That's my understanding just from very limited experience on the vfx side.

garrettia

thank u all for answer guys.