I have been trying for ages to get satisfactory results with subsurface (both
POV 3.7 and Samuel Benge's fastSSS). I imagine I am doing it wrong in some way
as I can never get the results that I want! So I got to debating with myself as
to the whichness of the why and tried using Irid instead. After various tests I
came up with the following, ramping Irid up (2), using low lighting (fade
distance) and a small thickness (0.05). The results are as you see (but be
advised these are all very much works in progress!)

One thing of note. It is very much faster than SubSurface. Also, I was wondering
whether there was any mileage in being able to use a colour component for the
strength, eg Irid <1, 0.7, 0.5>.

Post a reply to this message

Attachments:
Download 'test01.png' (752 KB)

Preview of image 'test01.png'

"Simon J. Cambridge" <nomail@nomail> wrote:
> I have been trying for ages to get satisfactory results with subsurface (both
> POV 3.7 and Samuel Benge's fastSSS). I imagine I am doing it wrong in some way
> as I can never get the results that I want! So I got to debating with myself as
> to the whichness of the why and tried using Irid instead. After various tests I
> came up with the following, ramping Irid up (2), using low lighting (fade
> distance) and a small thickness (0.05). The results are as you see (but be
> advised these are all very much works in progress!)
>
> One thing of note. It is very much faster than SubSurface. Also, I was wondering
> whether there was any mileage in being able to use a colour component for the
> strength, eg Irid <1, 0.7, 0.5>.

Number two! Just to add that these images are supposed to be 'portraits' though
I have tried this technique in complex environments as well (halls, rooms, etc).

Post a reply to this message

Attachments:
Download 'test02.png' (589 KB)

Preview of image 'test02.png'

"Simon J. Cambridge" <nomail@nomail> wrote:
> "Simon J. Cambridge" <nomail@nomail> wrote:
> > I have been trying for ages to get satisfactory results with subsurface (both
> > POV 3.7 and Samuel Benge's fastSSS). I imagine I am doing it wrong in some way
> > as I can never get the results that I want! So I got to debating with myself as
> > to the whichness of the why and tried using Irid instead. After various tests I
> > came up with the following, ramping Irid up (2), using low lighting (fade
> > distance) and a small thickness (0.05). The results are as you see (but be
> > advised these are all very much works in progress!)
> >
> > One thing of note. It is very much faster than SubSurface. Also, I was wondering
> > whether there was any mileage in being able to use a colour component for the
> > strength, eg Irid <1, 0.7, 0.5>.
>
> Number two! Just to add that these images are supposed to be 'portraits' though
> I have tried this technique in complex environments as well (halls, rooms, etc).

And number three! I also tried Irid on the armour. The jury is still out.

Post a reply to this message

Attachments:
Download 'test03.png' (468 KB)

Preview of image 'test03.png'

Am 30.04.2016 um 16:59 schrieb Simon J. Cambridge:
> I have been trying for ages to get satisfactory results with subsurface (both
> POV 3.7 and Samuel Benge's fastSSS). I imagine I am doing it wrong in some way
> as I can never get the results that I want! So I got to debating with myself as
> to the whichness of the why and tried using Irid instead. After various tests I
> came up with the following, ramping Irid up (2), using low lighting (fade
> distance) and a small thickness (0.05). The results are as you see (but be
> advised these are all very much works in progress!)
> 
> One thing of note. It is very much faster than SubSurface. Also, I was wondering
> whether there was any mileage in being able to use a colour component for the
> strength, eg Irid <1, 0.7, 0.5>.

Subsurface Light Transport and Iridescence are two entirely unrelated
phenomena in nature, and entirely unrelated mechanisms in POV-Ray.

If you find that the `irid` mechanism "kind of" does what you want,
you're likely in for much frustration if you try to coax it into doing
_exactly_ what you want, because it is probably designed to do something
entirely different.

Of course, figuring out what is _really right for you requires to get a
clear grasp of what you actually want to achieve.


SSLT is designed to make skin (and other materials, like wax, soap, or
even milk) look less "hard" than the results you can get from classic
rendering, to the point that it may look a little bit translucent.

I'd like to point out that your attempts at employing `irid` for more
realistic skin utterly fail in this respect (and inevitably so, for some
very fundamental reason); your people still look like made from some
hard plastic or plaster or something, and always will no matter how hard
you try with `irid`. There are other fast cheats that might do the trick
depending on the lighting conditions, but iridescence simulation is
certainly none of them.


If, on the other hand, what you're really after is just making the skin
look a little bit more "oily", then you're on the right track after all.
That's an effect that SSLT is unfit to simulate (so in that case it's no
surpsise you're getting nowhere with that mechanism), while `irid` is
indeed perfectly tailored for it.


As for allowing to modulate the `irid` effect on a per-colour basis,
that would be physically nonsensical, as iridescence _generates_ colour
effects, and it does so based on some very fundamental laws of nature.

Post a reply to this message

clipka <ano### [at] anonymousorg> wrote:
> Am 30.04.2016 um 16:59 schrieb Simon J. Cambridge:
> > I have been trying for ages to get satisfactory results with subsurface (both
> > POV 3.7 and Samuel Benge's fastSSS). I imagine I am doing it wrong in some way
> > as I can never get the results that I want! So I got to debating with myself as
> > to the whichness of the why and tried using Irid instead. After various tests I
> > came up with the following, ramping Irid up (2), using low lighting (fade
> > distance) and a small thickness (0.05). The results are as you see (but be
> > advised these are all very much works in progress!)
> >
> > One thing of note. It is very much faster than SubSurface. Also, I was wondering
> > whether there was any mileage in being able to use a colour component for the
> > strength, eg Irid <1, 0.7, 0.5>.
>
> Subsurface Light Transport and Iridescence are two entirely unrelated
> phenomena in nature, and entirely unrelated mechanisms in POV-Ray.
>
> If you find that the `irid` mechanism "kind of" does what you want,
> you're likely in for much frustration if you try to coax it into doing
> _exactly_ what you want, because it is probably designed to do something
> entirely different.
>
> Of course, figuring out what is _really right for you requires to get a
> clear grasp of what you actually want to achieve.
>
>
> SSLT is designed to make skin (and other materials, like wax, soap, or
> even milk) look less "hard" than the results you can get from classic
> rendering, to the point that it may look a little bit translucent.
>
> I'd like to point out that your attempts at employing `irid` for more
> realistic skin utterly fail in this respect (and inevitably so, for some
> very fundamental reason); your people still look like made from some
> hard plastic or plaster or something, and always will no matter how hard
> you try with `irid`. There are other fast cheats that might do the trick
> depending on the lighting conditions, but iridescence simulation is
> certainly none of them.
>
>
> If, on the other hand, what you're really after is just making the skin
> look a little bit more "oily", then you're on the right track after all.
> That's an effect that SSLT is unfit to simulate (so in that case it's no
> surpsise you're getting nowhere with that mechanism), while `irid` is
> indeed perfectly tailored for it.
>
>
> As for allowing to modulate the `irid` effect on a per-colour basis,
> that would be physically nonsensical, as iridescence _generates_ colour
> effects, and it does so based on some very fundamental laws of nature.

Thanks clipka.

I have tried with SSLT, I really have, but it never seems to give me what I am
after, which was why I went physically nonsensical. Skin ends up looking like
wax, or artifacts get generated (on meshes - light patches on vertexes).
Certainly, having played with irid, I do think there is mileage in the oily
effect, but only if you use it sparingly.

I have also tried scattering media and having objects inside objects, but that
makes modelling a nightmare. I would much prefer to get SSLT right as that is
clearly the most physically accurate, but I cannot seem to get what I am after
(artistically).

Post a reply to this message

Am 30.04.2016 um 18:20 schrieb Simon J. Cambridge:

> I have tried with SSLT, I really have, but it never seems to give me what I am
> after, which was why I went physically nonsensical. Skin ends up looking like
> wax, or artifacts get generated (on meshes - light patches on vertexes).

I know the artifacts you are referring to, and haven't found the time
and ideas to battle them yet; they should diminish if you subdivide your
mesh more rigorously though.

If you get a waxy look, you're overdoing it.

Also, in real life the SSLT effect is different for different portions
of the body; however, currently POV-Ray does no yet provide any means to
simulate this, except for averaging an SSLT- and a non-SSLT texture (or
multiple SSLT textures with different settings), which does not do full
justice to the real effect.

Post a reply to this message

clipka <ano### [at] anonymousorg> wrote:
> Am 30.04.2016 um 18:20 schrieb Simon J. Cambridge:
>
> > I have tried with SSLT, I really have, but it never seems to give me what I am
> > after, which was why I went physically nonsensical. Skin ends up looking like
> > wax, or artifacts get generated (on meshes - light patches on vertexes).
>
> I know the artifacts you are referring to, and haven't found the time
> and ideas to battle them yet; they should diminish if you subdivide your
> mesh more rigorously though.
>
> If you get a waxy look, you're overdoing it.
>
> Also, in real life the SSLT effect is different for different portions
> of the body; however, currently POV-Ray does no yet provide any means to
> simulate this, except for averaging an SSLT- and a non-SSLT texture (or
> multiple SSLT textures with different settings), which does not do full
> justice to the real effect.

I am amazed you are considering tackling it at all. It must be a nightmare. (My
developer days are over, I'm afraid - it was all beginning to make my brain hurt
a bit too much). (And thanks for the tip on subdivision. I will try that out,
though I think I gave it a go some time back and had problems still.)

To be honest I have tried various settings with subsurface and found it way over
the top, artifact city or no appreciable effect at all (but, as I said above,
this was some time ago. Perhaps if I revisit the problem with renewed vigour and
a smaller mesh?)

I think the most frustrating thing, though, is the time it takes to render. I
set a render in motion, go away, return ten years later and bang my head on the
table because something is still not right in the image. Which is why I am
following the cheating path as it makes me feel a lot less like Sisyphus.

Has anybody else used the Samuel Benge (thank you Samuel, btw) fastprox macros,
and had satisfaction, do you know? Is this a route worthy of further travail?

Post a reply to this message

"Simon J. Cambridge" <nomail@nomail> wrote:

> Thanks clipka.
>
> I have tried with SSLT, I really have, but it never seems to give me what I am
> after, which was why I went physically nonsensical. Skin ends up looking like
> wax, or artifacts get generated (on meshes - light patches on vertexes).
> Certainly, having played with irid, I do think there is mileage in the oily
> effect, but only if you use it sparingly.
>
> I have also tried scattering media and having objects inside objects, but that
> makes modelling a nightmare. I would much prefer to get SSLT right as that is
> clearly the most physically accurate, but I cannot seem to get what I am after
> (artistically).

Why don't you post your SSLT results for comparison? If you're willing, you
could also post your scene and so people can play around with it to help you
out.

Post a reply to this message

On 30-4-2016 19:45, Simon J. Cambridge wrote:
> Has anybody else used the Samuel Benge (thank you Samuel, btw) fastprox macros,
> and had satisfaction, do you know? Is this a route worthy of further travail?
>

Yes, I have. fastprox is very good although I prefer Edouard Poor's 
proximity macros because they seem more comprehensive to me and a bit 
faster over all. I have not used it for skin as such but for giving a 
wear-and-tear look to textures. It might be interesting to see what it 
does with skin though, an idea I shall pursue [takes ToDo list from pocket].

-- 
Thomas

Post a reply to this message

clipka <ano### [at] anonymousorg> wrote:
> Am 30.04.2016 um 18:20 schrieb Simon J. Cambridge:
>
> > I have tried with SSLT, I really have, but it never seems to give me what I am
> > after, which was why I went physically nonsensical. Skin ends up looking like
> > wax, or artifacts get generated (on meshes - light patches on vertexes).
>
> I know the artifacts you are referring to, and haven't found the time
> and ideas to battle them yet; they should diminish if you subdivide your
> mesh more rigorously though.
>
> If you get a waxy look, you're overdoing it.
>
> Also, in real life the SSLT effect is different for different portions
> of the body; however, currently POV-Ray does no yet provide any means to
> simulate this, except for averaging an SSLT- and a non-SSLT texture (or
> multiple SSLT textures with different settings), which does not do full
> justice to the real effect.

Just to say thank you for your advice. I went back, bit the bullet, threw
everything out the window and ... found three things.

The artifacts do indeed get better if you subdivide, but, and it is a big but,
the situation is vastly improved if you go assumed_gamma 1.0. You also get a
speed improvement. I had left gamma at 2.2 (my default - when ever I create a
new scene I cut and paste the environment - a possibly very bad habit). I hadn't
any idea gamma would make such a drastic difference. Ignorance, in this case, is
not bliss.

Oh, and my comment on using a colour component for Irid was meant in terms of
overriding the wavelength set in global settings for individual textures. It was
just a thought.

Post a reply to this message