Feedback on Workflow

CGStephanN

Hello polycount,

        I wanted to see if I could get some feedback on a workflow that I want to use to improve my 3D art. I know workflow is pretty unique per person/project but any suggestion is welcome. I'm wondering if I'm thinking of something the wrong way or if I'm missing steps, the worst part would be to go far along the process and then find out that the model is not good for animation and that is why I made the topic.

Goal

Conceptualize, and model a game character from start to finish from Maya to Unreal Engine. 

Details of Goal

• Humanoid character
• Polycount 25,000 - 35,000tris
• Needs Subsurface Scattering on skin
• Needs Ambient Occlusion 
• Animation Ready
• Optimized for Games

Workflow 

1) Base Mesh = Maya

2) Sculpt High Poly = ZBrush

3) Retopology = Mudbox 
Note: I think for this step I would use Automatic retopology, I've seen some good results though I am not sure how efficient it is on organic models compared to manual retopology. 

4) UV Unwrap = Maya 

5) Texture = Photoshop

6) Subsurface Scattering  + Ambient Occlusion + Normal Map = Zbrush

7) Baking (if necessary) = Maya

8) Rig = Maya

9) Weight Paint = Maya

10) Animate = Maya

11) Export

12) Import to Unreal Engine

NOTES & QUESTIONS
       Let me just say that weight painting and mirroring weights in Maya is the bane of my existence. I am not sure if it's as tedious in other programs but every time I mirror the weights they don't seem to go where they should even if the mesh is symmetrical. Any tips on how to improve weight mirroring? 

      Sometimes the models I make are very hard to deal with when rigging because they don't deform as I intended. Is this because it doesn't have enough polys in the bend area or are pro level models very reliant on helper joints, twist joints, and blend shapes? (I have tried different topology layouts for knee, elbows, and shoulders but they tend to make the model look weird when smoothed)

      Should I bake Subsurface scattering maps, ambient occlusion, and normal maps in a modeling software or should I do that within the game engine? I figured it should be in the game engine but maybe baking them would improve performance? Not sure.

      Hope this is not an obvious-answer question but, if I were to follow the standard edge flow of face modeling, am I more or less guaranteed to get proper facial expressions? The humanoid character is only a bit stylized, like an Overwatch character. 

     Should I convert the polys to tris within the modeling software or should I leave it to Unreal to change it for me? I've heard that most if not all engines take care of converting polys to tris. Or is it more of a control thing?

     

pior

Judging by the way you are describing your intended workflow ("I think for this step I would use Automatic retopology, I've seen some good results though I am not sure"" ; Baking (if necessary) = Maya"), tackling a complex 30k character at this time is probably a bad, bad idea.

Do not attempt to "figure out" a technical workflow while tackling a complex artistic endeavour at the same, it's the recipe for disaster and frustration.

There are indeed countless of different ways to get there. Some are incredibly convoluted and wasteful, and some are super efficient. So to figure out what works for you to create content indistinguishable from professional models and doing it efficiently, start simple. For instance, why note taking one of these guys through the whole pipeline :



Once you've done that you'll be ready for more complex stuff.

musashidan

CGSN said:

Workflow 

1) Base Mesh = Maya -

>>>Yes. You can create the final organised topology at this early stage and save the retopo step. This is a very     common approach in studios.

2) Sculpt High Poly = ZBrush

>>>Yes. Still good so far.

3) Retopology = Mudbox 
Note: I think for this step I would use Automatic retopology, I've seen some good results though I am not sure how efficient it is on organic models compared to manual retopology.

>>>No. Build an organised mesh topology manually. You need the practice. Personally, I am getting tired of hearing this from beginners. If you plan to take character work seriously then learn about organised topology. Alternatively, refer to answer 1, above.

4) UV Unwrap = Maya

>>>Yes. Back on track.

5) Texture = Photoshop

>>>Definitely still a viable option. Mudbox is also another, as you have the program.

6) Subsurface Scattering  + Ambient Occlusion + Normal Map = Zbrush

>>>You are confused about this. SSS in ZB? What do you think you mean by this?

7) Baking (if necessary) = Maya

>>>NO. NO. NO. Maya baker is outdated and isn't synced with UE4. There are many modern options now, all using the same modern tangent basis as UE4. I advise study on this.

8) Rig = Maya

>>>Good choice. Google the recently integrated rigging system that plugs directly into UE4. It's called ART.

9) Weight Paint = Maya

Y>>>es, again. Practice makes perfect.

10) Animate = Maya

>>>As good a place as any. Better than most.

11) Export

>>>This one you got right.....

12) Import to Unreal Engine

>>>This one too.

NOTES & QUESTIONS
       Let me just say that weight painting and mirroring weights in Maya is the bane of my existence. I am not sure if it's as tedious in other programs but every time I mirror the weights they don't seem to go where they should even if the mesh is symmetrical. Any tips on how to improve weight mirroring?

>>>As mentioned above, practice makes perfect.

      Sometimes the models I make are very hard to deal with when rigging because they don't deform as I intended. Is this because it doesn't have enough polys in the bend area or are pro level models very reliant on helper joints, twist joints, and blend shapes? (I have tried different topology layouts for knee, elbows, and shoulders but they tend to make the model look weird when smoothed)

>>>This is when you put time and effort into studying topology. See answer 3, above. Export characters from Epic UE4 example projects and study the wireframes. Check the WIKI here for good examples of organised topology.

      Should I bake Subsurface scattering maps, ambient occlusion, and normal maps in a modeling software or should I do that within the game engine? I figured it should be in the game engine but maybe baking them would improve performance? Not sure.

>>>There is a lot of confusion with this statement.

      Hope this is not an obvious-answer question but, if I were to follow the standard edge flow of face modeling, am I more or less guaranteed to get proper facial expressions? The humanoid character is only a bit stylized, like an Overwatch character.

>>>Again, study professional organised topology. There is no 'standard'. Just many variations.

     Should I convert the polys to tris within the modeling software or should I leave it to Unreal to change it for me? I've heard that most if not all engines take care of converting polys to tris. Or is it more of a control thing?

>>>No. Your mesh will be triangulated before baking. An important step.

     

My main advice to you here is exactly what @pior has already stated. It is very clear from your post that you are confused and misguided about a lot of the fundamentals. This will lead to frustration if you go ahead with a complex project like this. Start simple. Build upon your knowledge gradually. Learn the different steps/approaches in smaller, more manageable projects. A solid, efficient workflow can take years to build.

Your main problem seems to be that you are trying to rush it too much. This is understandable as I'm sure you're excited. However, it can lead to bad habits, a lot of wasted time, a haphazard workflow, and poor results. Slow things down. This stuff takes a lot of time. Endless hours of dedicated, quality study and practice.

Self-experimentation in small, byte-sized projects that focus on single aspects of the workflow/pipeline are key to progressing. I wish I had discovered this many years back in my own self-taught(for want of a better phrase) journey. CG is constant learning. Even after a decade the learning continues. 2 decades.....3...... the learning never ends. So why rush?

CGStephanN

Thank you @musashidan and @pior , I really appreciate all the feedback. I'll go ahead and do as you both said about micro projects.  

NoRank

I'm kinda curious, where did you hear that using automatic retopology is a "good" alternative, specially for characters?

It's weird, this is like the third time only this week that someone has this idea and I'm here wondering why, from where did you guys got that info?

Don't take this as an offense, I'm really curious since you're a beginner and most of the time when you're a beginner you actually get your info from other sources, which makes me wonder who or what is the source of this.

CGStephanN

@NoRank No worries. To answer your question, I can't really say I got it from one single place, I just "deduced" that automatic was the way to go based on a couple of YouTube videos that did it, though they weren't necessarily about making a game character from start to finish. Seeing ZBrush as such a powerful tool after mostly using Maya, I suppose it's just a thing that comes with being awestruck with ZBrush  I admit that it was a pretty naïve assumption considering that we have software that does "automatic" UV Unwrapping and I know full well how well that goes.

johanflod

I think it is very hard to give feedback on this. Please show some concepts of what you want to do. 2D Concepts that covers the whole body.  Then on top of that you need some photo references( to use when you work in ZBrush) of each individual item that the character wears (even if you go for a carton style). You will quickly notice that this takes time to gather. 

I strongly agree with pior, start with something simple to test the pipeline. Make a basemesh of like 1000 polys and take it through all the steps. To skin something which is 1000 polys is relatively quick. To texture a 1000 poly character is also relatively quick.

To fix skinning/texturing/bake problems on a 25,000 - 35,000 tris character is a different beast. 

I would also suggest that you keep the texture size small for the first pass. Like 512 textures or something like that. Just to get the big shapes right to start with. 

Once you have the 1000 poly character all the way through you have made some very important findings. 

Each area of your character pipeline is a potential time sink. If you are new to ZBrush you can easily spend months there for example. 

Doing the basemesh for highres modelling is quite different from the mesh that you make for skinning. 

basemesh for high poly = proportions are right and the mesh will subdivide nicely.

basemesh for skinning = the mesh has more edgeloops in areas that deform a lot. More geometry in face, hands, knees, shoulder, hands etc.