Topology before sculpt or retopo after sculpt?

Yesterday I finished my base mesh and am wondering what to do next.
Should the sculpt base mesh already have a good topology or can I just keep the quads evenly distributed and use retopo after the sculpt?

I was thinking about the following workflows:

  1. Base mesh (Box modeled)
  2. Sculpt
  3. Retopo
  4. Unwrap (The retopo-ed)
  5. Bake (From sculpted to retope-ed version)


  1. Base mesh (Box modeled)
  2. Topology
  3. Sculpt
  4. Unwrap
  5. Bake (From high multires level to low multires level)

At the moment I am unsure which one would work best.
Hopefully someone can give me some advise about good workflows.

Either work.
If you want to be fluid and creative, not be restricted to any fixed topology you can:
1 Base mesh (or even just a sphere, subdivided cube or very low poly base mesh)
2 Sculpt
3 Retopologise
4 Refine sculpt using new topology
5 Unwrap
6 Bake

How exactly can I refine the sculpt? Should I use Reshape to copy the details over to the new retopologised mesh?

Thank you. Will just wait and see for myself as soon as I am at that step :wink:

I think a lot depends on personal preference and the nature of the model. I would be more likely to use the second workflow, as I like to have as much control over final topology as possible. Retopo’s “automated” nature may not always produce an optimized mesh, though undoubtedly it would be useable. Defining the topology from the get-go also helps me envision such things as the best UV seams for the model, something Retopo may not be able to optimize for.

But that doesn’t mean it’s the better way to go, just the one I’ve found more useful. I don’t find it hinders creativity or a “fluid” workflow at all.

But the retopo tool does not seem to be all that automatic.
For example have a look at the following tutorial:

As you can see its mostly done by hand and gives you the tools to redefine the topology just the way you want by just drawing on top of the high poly mesh. And of course. At the end it’s just about preference and the kind of model that’s being worked on. For hard edge / non-organic models the second work flow seems to be more logical.