Home Technical Talk

Super fast bakes - aka - teodar's baking method

sublime tool
Offline / Send Message
teodar23 sublime tool
Hi,

I would like to share a bit of info about my process of baking (no, this is not a cooking tutorial) which some of you might already be using, or at least, be aware of - at least as a work-around.
First off, this is not my usual baking process, its something that I find faster in some scenarios (eg: hard surface props) and as any process has its pros and cons.
So here goes:
I usually start modeling a 'mid poly' model that can be easily converted to a low poly by removing some edges or verts. This mid poly model can also be turned into a high poly by adding some support loops and if needed a turbosmooth. I say 'if needed' because depending on the model or model part you can sometimes get away with just a chamfer modifier on top of the 'mid' or low poly model.
Here's an example:
OK, so we have the low version all cleaned up and UV mapped and the hi-poly (which is the low + support loops). Now we can bake, right?
Wrong! If we take a look at our hi poly we can notice some shading artifacts:

These are minor to some, but if you want that perfect bake, there's a solution.
Obviously the smoothing issues are created by the extra geometry introduced by the boolean operations that were performed when creating the model.
No worries, we still have a copy of the model before boolean ops and we can also use primitives (cylinders in this case) to bake proper smoothing back into the low poly. All we need is a mask to blend between the 3 normal maps (A, B and C). We can create such a mask by generating a UV islands map or - what i ended up doing - a vertex color map that is generated from the high poly model. The vertex colors were as follows: blue for where i want to use the normal map from the non-boolean or exterior cylinder (B), red for the interior cylinder (C) and black for the high poly normal map (A). In retrospect this could have been done better but whatevs.
This seems like a lot of work but its not. For this model it took ~30 secs to bake all the maps in Toolbag and a couple of mins to fiddle in Photoshop.
After blending the three normal maps this is the result:


Of course the gain is minimal and the artifacts can be ignored in most cases. Of course there are better ways of creating that perfect bake.
But i find that this method is fast and less of a mess than other traditional ways.

Anyways, i hope you enjoyed this small hack and maybe you can share other similar tips and tricks that are somewhat weird but helped you in the past.

Replies

  • musashidan
    Options
    Offline / Send Message
    musashidan high dynamic range
    So you need 4 meshes and 3 normal maps for each model? So if you had an asset with 30 seperate parts you'd have to manage 120 meshes and set up IDs for all 30? Seems like a lot of work for a result that can be achieved by placing some strategic n-gons(or fwvn/normal transfers in a lot of cases) on the high poly to clean up the shading.

    Each to his own though. :)
  • teodar23
    Options
    Offline / Send Message
    teodar23 sublime tool
    As i said in OP this flow doesn't cater to all scenarios.
    For a more complex model or a lot of models, some tasks can be automated, like a macro in PS for blending the normal maps.
  • Prime8
    Options
    Offline / Send Message
    Prime8 interpolator
    Interesting idea, but not super fast, prone to error and only usable situational. That example could be fixed easily with weighted normals.
  • Cathodeus
    Options
    Offline / Send Message
    Cathodeus polycounter lvl 14
    Yeah "Weighted normals" are you friends. or force face normal to poly.
  • Neox
    Options
    Online / Send Message
    Neox veteran polycounter
    a very convoluted approach, this would be faster already when you would have prepped the base version a bit differently. I think this is stuff people avoid automatically after the modelled the N-th version of this case.
  • Cathodeus
    Options
    Offline / Send Message
    Cathodeus polycounter lvl 14
    The good thing is that you found a solution yourself. And that is good [and the approach while being complex is not unintresting].
  • teodar23
    Options
    Offline / Send Message
    teodar23 sublime tool
    Yes FWN are awesome but i find that flow a bit destructive and cumbersome, at least in 3dsmax.
  • musashidan
    Options
    Offline / Send Message
    musashidan high dynamic range
    teodar23 said:
    Yes FWN are awesome but i find that flow a bit destructive and cumbersome, at least in 3dsmax.
    You're right, it is, but there are some good scripts to help with that. Also, the Max devs are aware of it and I'm sure they'll address it sooner than later(hopefully).
    There are so many tools and workflows these days to make our baking life so much easier than it once was, I suppose that's why people are wondering why you would put yourself through all of that when there are much more straightforward ways to do it. :)
  • cptSwing
    Options
    Offline / Send Message
    cptSwing polycounter lvl 11
    Please tell me who to bribe for a 3ds max FWN modifier. I can't believe we're still discussing this in 2019.
  • teodar23
    Options
    Offline / Send Message
    teodar23 sublime tool
    ...that's why people are wondering why you would put yourself through all of that when there are much more straightforward ways to do it. :)
    It wasn't planned, i don't know why it came out like that, but this is just a workaround for the situation that i was in.
    The model is done with a shape + lathe modifier so for the high poly i could have easily amped the radial sections for the lathe and get a much cleaner model for the bake. But that would require going back a few steps.
    Either way thanks for sharing your views on all of this, i am somewhat new to this.
  • Kanni3d
    Options
    Offline / Send Message
    Kanni3d ngon master
    Neat way of blending multiple results into one perfect one, but seems a bit too convoluted as someone mentioned.

    I get it's personal/portfolio only, but these sorta habits would never fly in production. I always think about "what if someone on my team needs to use my mesh and doesn't know that one hacky fix that I came up with?'", or, "what if i get back to this later down the line because I noticed a design flaw/error/etc' and I would have to go through all those steps again just to bake out my normals...

    I'd much rather spend the time & time to work the topology to get a proper looking high-res mesh that bakes perfectly fine in one go. :smile:

Sign In or Register to comment.