Download

Question on Reducing High Level Mesh Complexity

Hi All,

Some of the resources you find on sites like Turbosquid have huge poly counts. Like this one:

Naturally, it’s because they look fantastic. However, I’ve never reduced anything with 400k poly count down to more like 2k and I’m wondering:

  1. If it’s even realistic to buy and reduce the complexity of something with these many poly’s into a game environment?
  2. If so, can you maintain decent quality ?
  3. If so, what’s the best way to go about reducing the complexity to get it to a good spot?

Thank you!
Josh

You would need to do it by hand, retopologize and then bake the textures.

Had to deal with some of the CAD models and a lot of architecture. In most of the cases we had to re-model most of the geometry, not really because of the amount of polygons but it’s just close to impossible to get proper UVs. At the same time you can do it much faster on top of the existing model than research for reference images or model against cad drawings.
Cost wise it makes sense to go with such approach if you want to save time as those 200-300$ would rarely get you a good quality model if you outsource.

Ive been using adobe’s decimator haha, it does very well imo.

https://www.mixamo.com/decimator

Got any recommendations?

Think it could handle 200k polys down to about 2k well?

This thread talks a bit about taking a bought mesh not designed for games and making it game ready https://forums.unrealengine.com/showthread.php?109152-Lets-create-a-complex-lightmap-Open-discussion

I used it with Fuse chars, went from ~60k down to 4k and it looked pretty **** normal, and that’s a person. i think it might be able to do it quite well depending on the exact type of model, its free so its worth a try if you happen to have any test models

Thanks! I’ll try it out and report back.

As I said, if you want to save time, model proper low poly using high poly as reference. You can snap to vertices or shapes to speed up the process. Some parts you will have to throw away completely to stay under the budget and do them as a texture. Something that works for a character recently won’t work for a warship as a lot of original polygons are not in the man shape but smaller details. But give it a try, mayb Decimator will get you good enough results.

Doing manual retopology and baking the rest into textures will always get you better results in the end than a “decimator” of any kind. Especially when starting with very high poly models and you want the end result to be game ready low poly. They are great and save a lot of time in some cases, but whenever you try to reduce the polygon count by a large percentage, it will go bad pretty easily. Once you get the hang of it and figure out your workflow, manually doing it becomes rather simple (if a bit boring and somewhat time consuming).

For a lot of these models all you would need to do for the low poly is take a cylinder, sphere, or cube, add a bunch of subdivisons, and shrink wrap around the highpoly model, and cubes/spheres/cylinders would be much easier to unwrap than a decimated mesh.

Reducing the count is not all the difficult were if I need to drop the count pro-optimize in 3ds Max does the job but the problem with Turbo Squid is not as to usability but the application of a Royalty Free License as to what you might consider as “NOT” being fair use.

for example

e. No Obscene or Unlawful Use. You may NOT use Stock Media Products for any defamatory, harassing, pornographic, obscene, or racist purpose, or to infringe any party’s Intellectual Property rights.

That is a red flag as to fair use that is non-descriptive as to what would be considered a violation of “this” licensing requirement that crosses the line of enforced censorship.

P.S. Rule number 1 “always” read the terms and conditions of the licensing requirements.