Hair planes, how come they always look like dreadlocks?
vertex
SO I've switched from fibremesh to hair planes because marmoset can't render fibres well, and it's less expensive.
I decided to go by one of blizzard's references, some un-rendered artwork of the druidess of teldrassil, from the vanilla cinematic.
Fairly thin planes, but for some reason their's didnt have this seperated.. dreadlock look.
I have a lot less strands than they do, but I figured just layering it with hundreds of these isn't gonna fix it, so what will? I've tried all kinds of different texture shapes, nothing seems to work.
The only other way I can think of fixing this would be having larger planes, or merging planes.
I decided to go by one of blizzard's references, some un-rendered artwork of the druidess of teldrassil, from the vanilla cinematic.
Fairly thin planes, but for some reason their's didnt have this seperated.. dreadlock look.
I have a lot less strands than they do, but I figured just layering it with hundreds of these isn't gonna fix it, so what will? I've tried all kinds of different texture shapes, nothing seems to work.
The only other way I can think of fixing this would be having larger planes, or merging planes.
Replies
at any rate the planes seem awfully thin, at least mine always tend to overlap and cover the character a great deal more. probably not helping with the 'dreadlock' look right there.
but the head isn't posed.
There's texture preview?
The guide I follow said the polypaint it in zbrush, but that would mean I have to subdivide. Too thin? So just make them wider then? I'll try that tbh.
also set tool -> display properties -> double if you haven't already.
like so -
personally, working in Z to tweak hairplanes beyond just general volume tweaks sounds like a really bad idea to me. especially for long hair it's so much more convenient to use curves/splines in an actual 3d app and keep everything tidy and separate in layers and groups. working in Z at that stage sounds like the result will be uneditable polysoup.
Sculpting hair, works for stylized looks with the volume of hair being defined by specific shapes. Like: https://ihaztoys.com/ for example. Amazing hair, but defined by big volumes, not made from groups of cards.
Creating cards first and placing them, leads to a loss in overall shape, volume and cohesion. With the cards first approach, It's hard to get big shapes that define a hair style.
With a lot of cards over lapping and criss-crossing it ends up looking like spaghetti head and can lead to a lot of overdraw issues, if you don't carefully manage how many planes are stacked underneath one another.
There is fast sorting which treats the transparent pixels as on or off but doesn't look very good because it isn't transparent. There are some techniques that you do to fuz it out and smooth it but those have their own overhead and require some tech work to get working correctly with whatever engine you're using.
So you want to carefully curate how many transparent pixels you are putting in your cards. Your first screenshot has a lot of cards that have a few polys that are completely transparent, you can get rid of those polys and save yourself some processing.
Nvidia HairWorks and TressFX have been around for a while but games are finally get to the point that they are throwing enough tech into hair to get these systems up and running. That doesn't mean those are the only tools you need to learn, those are mostly middleware to get your awesome hair into the engine, how you build that hair is mostly up to you. I would look to film studios and what people like Johan Lithval are doing. You want to be building a portfolio that speaks to what studios will be using on future titles, not what they released 5-6 years ago.
Games like The Witcher 3 and Horizon Zero Dawn are leading the way but there are a lot more studios doing it and not all of them are huge. Almost everyone working on future releases unless it's for really low end hardware are mixing splines and planes and doing work similar to traditional hair workflows that you see in films. They rely on a mixed bag of techniques to make it preform well-ish, aggressive LOD, tessellation, various baking techniques and some complex shaders that get fine tuned for their specific technical process.
Tools like GMH2, help convert geometry to hair and then hair to cards. XGen in Maya helps the hair creation process and it can go much deeper than what they show in the Witcher video, which is actually quite involved. It probably seems like a lot of learning and experimentation but since you're at that phase I urge you to check it out rather than chase down a rabbit hole that leads to a dead end. There isn't one particular workflow or one singular tutorial that I've found, that is "the bestest way ever" but that is the direction that the industry is trending so I would check it out. It's not easy, its actually REALLY hard to make looking hair, especially hair that looks good and preforms well. But you are at that phase where it could pay off to chase a new direction.
As you scratch around you'll probably find better ways to work and you might even blaze a few trails
in my experience this is handled by LOD's quite easily. meaning only your closeup model actually receives the full treatment and just a few meters away fro the camera you can swap in a far more optimized mesh that can use any combination of solid underlayer and thin veneer of hair cards. it's not like hair cards cast nice shadows with the typical lighting fidelity in an actual level nor do they behave well when mip mapped and texture filtered or subjected to screen space post-processing.
pretty quick to do such an optimized model once you have shape, volume and textures to bake from.