How to make my subd topology more low-poly for real-time rendering?

OblongUV

This model will be used for real-time rendering so I'm trying to optimize it more. This is the low-poly, but the high-poly is just a linked object with a subd modifier (Blender). It is full of edgeloops that are just there to make the low-poly and high-poly shade better. (Marked in green are a few of them). Do I actually need these edgeloops that are just there for shading, or is there a way I can unlink the two and remove those loops and use custom normals instead on the low-poly? It's my first game-asset so I'm unsure how you go about it. If I just unlink and remove them on the lowpoly, the shading is much worse after bake. I assume that even if you bake from high-poly, the result is still somewhat contingent on the low-poly normals?

Eric Chadwick

Hard to say without knowing the context. How big is this piece, and how close does the player get to it? 

Generally, we try to optimize away any loops which are not adding to the silhouette. 

Only judge shading in the context of the actual game engine, and preferably in the game level where it’s lit. 

After a while doing this, you’ll start to learn how and why certain topologies work, so you can do less in-game viewing and more editing in the modeling app. But without that experience, it’s best to view in-game more often.

Some tips here: http://wiki.polycount.com/wiki/Normal_Map_Modeling

gnoop

Use  decimate modifiers in Blender and vertex groups to set priorities and keep some loops necessary for animation intact .   or multi res in MAx  or simplygon .   They will try to keep  shading up to certain  percentage  but to get best result use Blenders  data transfer modifier to project vertex normals from original mesh to decimated. 

Blenders decimate  is the best IMO  btw.

Or just slide vertexes manually  with auto merge option on  and data transfer on top of this to keep normals  .

Then for next Lod and so on .

OblongUV

Eric Chadwick said:

Hard to say without knowing the context. How big is this piece, and how close does the player get to it? 

Generally, we try to optimize away any loops which are not adding to the silhouette. 

Only judge shading in the context of the actual game engine, and preferably in the game level where it’s lit. 

After a while doing this, you’ll start to learn how and why certain topologies work, so you can do less in-game viewing and more editing in the modeling app. But without that experience, it’s best to view in-game more often.

Some tips here: http://wiki.polycount.com/wiki/Normal_Map_Modeling

gnoop said:

Use  decimate modifiers in Blender and vertex groups to set priorities and keep some loops necessary for animation intact .   or multi res in MAx  or simplygon .   They will try to keep  shading up to certain  percentage  but to get best result use Blenders  data transfer modifier to project vertex normals from original mesh to decimated. 

Blenders decimate  is the best IMO  btw.

Or just slide vertexes manually  with auto merge option on  and data transfer on top of this to keep normals  .

Then for next Lod and so on .

Thanks gonna check it out