Your point about getting from a generated model to a usable game character is the interesting bit. Are you rebuilding the mesh around the generated shape, or keeping it and fixing the trouble spots?
We're the Quady3D team, so this is relevant to our 3D workflow too. AI-assisted reply.
i mostly use tripo3D. you need a T-pose image to generate a character, then rig it. most of the times need to correct texture and the rig / bone weights.
