We used a lot of apps to finish this at my studio. Sculpted in Zbrush. Blender played an important role for various modeling and retopo jobs. And we were going to render it there but Blender could not handle all of the data so we rendered it in Maya with Arnold
Apps: Zbrush, Substance, Painter ,Mari, Blender Maya.
sculpt turntable
Holy Diver, the details on the cloth. Tres cool.
I featured you on BlenderNation, have a great weekend 
Thanks! Much Appreciated.
You’re on the #featured row! 
Cool. Thanks!
And yes. Insane. This guy is the best in my studio. I think there was something like 2 months work in sculpting.
But finishing her was a team effort. At the time I was training interns from the local uni. For three artists there was over a months work full time to retopo texture and rig it. If I can get some time I will try to render some motion.
Needless to say. Those three interns got hired.
Here are some of the original concepts. But I always let my sculpting artist pretty much take the reigns from there and take it to another level. Though for the final we do try to stick to the color theme.
This character was based on my original concept which is here:
Argument from the team was we did not want to make her too ugly because in the story she had a boyfreind… hmm… well, I conceded. Maybe I can use her someplace else.
Well she sure does have razor sharp lipp gloss 
Wow the level of detail is impressive. Could you let us know how many polys? Curious to know as you said blender could not handle it.
Neat. Mind the topology on the hands, for instance. Every knuckle has a grid fill showing mesh curvature. This is reference-level. Manually done? Or automatised?
Almost the entire model was hand retopo. Most of that with Maya retopo tools.
Some of that “retopo” was re-creating the chains, beads, rope and so on with Blender array modifiers and curve modifiers. Blender made quick work of re-creating the Zbrush artist’s arrays. This way we could uv map smaller segments and save time and image resources.
Some cloth parts were Zremesher.
The final model was assembled in Blender from all of the various parts. As we were going to render it in Cycles.
It was the images that capped it off in Blender. Arnold handles image data completely differently. I am not sure it was totally due to polygons. Anyway this is a known limitation of Cycles. At least I know it now… lol. But I have ran into other people here with the same issue. Cycles needs to get better handling of image data. We had a lot of UDIMS from Mari as well which we had hacked a way to make work in Blender.
But it was all for not since Blender can’t render it. I brought this up in the UDIM thread. Very important improvement needed for Cycles if we are going to get UDIM.
May I ask why you needed so many images? To me it looks like large parts could be more easily done procedurally.
Great work, congrats…
Thanks! Appreciated.
Its not that many images for a hero feature film ultra-real asset like this. I would say we use less than half as what I wanted. And I started out being extremely conservative.
Regarding procedurals we are using those all over the place. It is not about using procedurals or not it is about using them within a painting app that gives you layers of control interactively with masking and layers as well as brushes in an app such as substance painter. Procedurals alone are not that useful for an asset like this. It is about a seamless use of proceedurals and painting interactively. The trade off off saving proceedurals till later is simply not worth the loss of interactive texturing workflow.
Blender needs to step up to the next level here. This is how assets are done. And Blender needs to support that.
I meant why not make the procedurals in Blender and only use the Substance images where needed but I guess I can see how that can be impractical.
Yeah, I got it. And definitely thanks for the suggestion. That is it essentially. We’d have to set up masking and material nodes in Blender. Nothing wrong with that workflow by itself other than that it can not integrate with the asset in Substance without a split up workflow that is far more tedious and slow than just doing it in Substance. Then there comes the challenge of making the asset portable to other apps and render engines. In which case of course we bake the procedural, but then we are back to images again. And possibly recreating the node set up in some way. Generally speaking images become the main input for material nodes no matter where we are rendering.
I am planning to port the asset to Unreal at some point in which case the whole thing is going to have to be refactored. That will be interesting fun…
I’ve dated worse . Seriously great character work and the rich detailing in the clothing is superb.
No surprise it took 2 months studio work. The detailing is crazy. But it has to be that way if it goes on the big screen…
Let’s call that’s a pain(t) point.
