usually I sculpt heads (almost every sculptural object) in orthographic view in ZBrush. but most of times, the final in-game character will not look like original concept, even if highpoly model was exactly same as concept. I think that is because of difference between orthographic and perspective view. am I right?
head concepts are usually a perspective view of a character and when I sculpt a perspective concept, in an orthographic view, it will be wrong when I look at it, in perspective view. (I think!)
this is my final sculpt in ZBrush:
and this one is his lowpoly and perspective shot in 3ds max (FOV=70):
now he is another guy!
did you have any similar problems with in-game characters?
and do you sculpt characters in orthographic views or in perspective?
(and such a stupid question: when I add images as attachment and put them in my posts, it will show a link, not my attached images. why?! :poly136: )
Replies
Its more like how you will see it in the final game/movie, so is more relevant.
I think it can be useful to work in both views. It's good to switch to perspective view once in a while to check on things, but having a lot of perspective, especially when working on a head, can be disconcerting. A lot of portrait photography is taken with more telephoto ranges to reduce unflattering distortions of the face, and trying to mimic that when modeling can make it easier to judge forms (for me).
Does the head model in Max look better when you zoom out and view it from a distance?
yes, it will be better but still he is another guy!
it's little strange to sculpt in perspective view in Zbrush! "F" will not fit model to view, camera will not rotarte around my last sculptes area...
now I'm tring to get same FOV as max in zbrush. differe between 0-180 in focal angle will not affect much on my model, but it will affect much in ZBrush deffault models! (rihno or dog) *shrug*
I found how should I mach ZBrush FOV with 3ds max.
the point is to put your model in global 0,0,0 and then use it to ZBrush. (I used GoZ for this)
and as your model is far from 0,0,0 the perspective strenght will be reduce.
no matter how big is model, FOV just depends on posotion.
shots are with 70 FOV and last one is orthograph mode:
everything works fine.
thank you all.
Bear in mind that if you're using photo reference, the camera may have a different field-of-view, so having a "high" perspective distortion setting in ZBrush may actually work against you in some cases, if you're trying to match a photo.
However, using straight orthographic view when working from references tends to lead to wonky proportions, in my experience (not just from me sculpting, but from seeing other people sculpting in ortho views too).
The only time I really use ortho views in ZBrush is to do precise axial moves/scaling or transpose positioning.
it was wrong habit to sculpt in orthographic view..
http://www.lesjones.com/www/images/posts/stepheneastwood-tile1.jpg
This is something I always struggles with, even with games...
I also have tried to sculpt in ortho mode in zbrush but I always find that because our eyes are used to seeing in perspective I would try to force perspective on the model without even realizing it. Resulting in a very strange looking head when I finally did put it in perspective.
This seems to make sense to me because a focal length of 35mm is (I think?) a very common focal length in cameras. since we're used to seeing portrait photos taken with a 35mm lens it seems natural to use that when we render. After all it's still called a "camera" in CG even if it's in a game or in a 3D app
Just remember not to place the camera too close to your character though, as you can see in the picture frankie posted, the closer you are to your object the more distortion you will get.
also keep in mind to check your model every now and then in a real 3d package, as zBrush doesn't have a real perspective, everything can break once you move it into an actual 3d app. This can get pretty annoying when doing likeness.
for presentation, it's always nice to use a lens photographers use for portraits, which is usually (take this with a grain of salt) 85mm and higher.
artquest has the right idea I think. The general consensus seems to be the human eye focal length is roughly 17-50mm. So I'd say aim for around the 35mm mark. Best thing to do is regularly check the model in your engine/3d package of choice as well.
http://www.photofidelity.com/blog/50mm-eye-fov-busted.html
http://www.clarkvision.com/articles/eye-resolution.html
http://hypertextbook.com/facts/2002/JuliaKhutoretskaya.shtml
So what can we take away from the above?
---
tldr; Even with a fixed focal length, you get varying degrees of persp distortion anyway. Traditional sculptors and portrait photographers like stepping back/using longer focal lengths/lower FOV because of lessening distortion. Maybe we should too.
The only safe practice is to regularly export to an ordinary 3D package, but that really is a pain because it means working blind half of the time! I have been using DeepExploration for that lately. One can also use GoZ with Max, Maya, Sculptris, and so on. However the best solution is still to export to Mudbox, since It will let you use tools similar to Zbrush to further tweak proportions.
I too really like the look and feel of an ortho bust - it gives that nice boxy kind of visual strength. It might actually be quite fine for some uses, like a small scale toy print since we basically see those "from a distance". But for game stuff it really doesnt work. I am not sure which FOV numbers work well, but in general, the default settings of any regular 3D app are quite representative of an average game camera. Now of course the scale of the model is important too. I found the default Mudbox cube to be a really good yardstick for a human bust.
A setup I've been using :
Good luck !
The FoV can always be dialed back (although it doesn't seem that far off to me), and there are other factors to take into consideration as well such as if local transforms mode is also active with perspective, or if you have it set to auto adjust to the distance in newer versions.
mudbox has a perspective, zbrush doesn't
Here's my 2 cents:
My personal method is to first guess perspective of the concept, then work in that perspective in zbrush. If I think the concept is drawn very flat, I use less perspective and if its very distorted with fov I use more. I usually tweak the fov several times in zbrush as I get further in the sculpt as well. 45 or 50 in zbrush is usually what I use for heads.
So once I get the model looking very close to the concept in zbrush, with major forms, then I export to keyshot for another comparison. This almost always leads to my model looking way off in multiple areas once in keyshot (keyshot is trustworthy with its fov distortion imo so how it looks there is all I care about). For heads, 85 focal length in keyshot seems about right usually. Then I'll take screen grabs from keyshot and overlay the concept in photoshop to see the differences. I then make the necessary adjustments blind in zbrush based on this back and fourth process of comparing concept to model in keyshot. You can spend a lot of time matching stuff up in the phase, but its almost always worth the effort. When I look at the model from before this pass it usually looks pretty shitty by comparison.
I find that larger objects matching in zbrush works better then smaller, so head studies are probably the most challenging in terms of dealing with zbrush's wonky fov.
This process yields the most accurate results for me, if anyone else has more advice on the subject I'd love to hear it!