ever
Scholar
- Joined
- Nov 13, 2008
- Messages
- 886
From my experience the artist usually creates a heightfield ( the darker a pixel is the higher it is ) by hand and then there's a tool ( crazybump is what I remember being used ) to transform this into a normal map.MetalCraze said:@ever - don't they originally make a high-poly model anyway and then make normal maps to use with low-poly one based on it?
High polygon models usually cost quite a lot to make. They require a lot of artist time to get something more from those extra vertices. It's why the work has become more segregated these days, you have one guy just focusing on making really detailed noses for all the characters while another guy works on the mouths, another on the legs and so on. Imo you get better stuff when each artist is sort of responsible for the whole character from head to toe.
To understand the difference in cost imagine a normal map for skin bumps created by taking a photo of cracked earth, inverting the colors for the heightmap and then running a tool over it to get the normal map. The texture coordinates are then just projected onto some character's arm using any common modelling software, and tweaked. Later on the engine will transform this into a million polygons by tesselation. Now imagine telling the artist "model that by hand".
In case you are wondering "why not use the height map directly for tesselation" it's because there's no way to calculate the normals in the domain or hull shader unless you like the look of flat shading