Is there some kinda formula when going from high to low poly? Like say you had something with 1 million verts or 25,000 verts. How low would you go, LOL. Searched online ,but really couldn’t find an answer.
Dulicate a( shift + D) the high poly, move a copy to a hidde layer for later.
Use vertex slide and set to 0 and hit enter
Then in the vertex slide data editor thing in the tool panel bottom do correct UV,
Now using vertex slide, edge slide, merge faces, join vertex with edge etc, you can retopo by hand.
Remove unesseary edges leaving only faces that preserve the ability to animate the rig as well as preserve the silhouette, for bakeing it is best if it almost all quads where you can pull it off,
Now create a new uv map on the low poly actor, and bake the normals, and other textures 1 at a time over to the low poly UV.
Another method is called cage mapping, where you operate on a mesh using subdivide simple x2 and subdivide smooth x1 and don’t apply them,
At the end you bake the subdivided mesh to the unsubdivided copy of the same mesh (normals mostly)
Lots of videosout there. There is no magic formula…you need to understand the process and work hard.
Or just use the poly reducer script in blender , which I don’t recommend.
I prefer to make the low poly mesh first and modify that to get the high poly. Then you don’t get ugly overhangs or backfaced normals too much. But I don’t do sculpting only hard surface.
As low as possible while still maintaining the overall shape (not the details). If you had a cube with 1mil vertices of detail, you would only need a regular cube to bake the data to (providing the UV’s are optimized and the image size is big enough).
The “formula” is pretty simple:
- as much details as visible
- as much fluent animation as possible
(the second one means … that the audience does not notice lag)
What are the limits?
Min details: A screen of one color.
Max details: Pixel size. You do not need more detail on elements that result in a single pixel. It will not be visible.
Min speed: One frame.
Max speed: In theory infinitely. Practically it is how much your eye and your brain can recognize as difference. Technically what the monitor can show.
You need to find a balance between visible details and display speed. It is somewhere between the above limits.
Remarks
In a 3D world with perspective view objects near to the camera are rendered bigger. That means you need more details on near objects. It also means you need less details on objects with greater distance (that is where LOD tries to help).
Golden rule: no details less than one pixel.
The display speed depends on processing speed. As much things to process as longer the creation of a frame will take. As longer this time as less speed you get. The goal should be to decrease processing time. Processing time is not just render time as the game usually does more than just render (physics calculation, logic processing …). This all adds up. Some processing can run parallel. If it does the longer time counts (but it does not add).
There are other aspects to keep in mind e.g. creation effort. Typically when you want more details you have to spend much more time into creating this details. This will dramatically increase your development time and costs. So you need to find a balance here as well.
It often is worth to spend the details on objects that are in focus of the audience such as the main character, while objects with minor priority get less details (e.g. a tree, a stone, or a background character).