An image I've been working on for a few days.
Thanks to Chris B for pointing me to the SweepSpline macro, and thanks 
to Mike Williams for actually making it. The handle is a swept object, 
if it wasn't obvious.
Everything visible in this image is a superellipsoid, prism, lathe, or 
swept spline, though a few cylinders were used for CSG.

As for the realism, the inside of the bell is too reflective but I 
haven't been able to find a good way to make it dull/unpolished but yet 
still look like clean metal. My attempts at micro and macro normals 
(using variations of the textures that Ive posted in the metals.inc 
thread a while ago) didn't work well, and there are very fine lines 
between completely black, way-too-bright, and fuzzy.

Render time 2:06:31 on a quad-core 2.6GHz, beta33. Minor focal blur, 
radiosity with decreased brightness, and area light used. I think the 
line up the middle of the inside of the bell is a seam in the HDR probe 
(which is St.Peter's Basilica from Paul Debevec's gallery), I will 
rotate it once I have more changes to make.

CShake

Post a reply to this message

Attachments:
Download 'handbellmk2.jpg' (75 KB)

Preview of image 'handbellmk2.jpg'

CShake schrieb:
> As for the realism, the inside of the bell is too reflective but I 
> haven't been able to find a good way to make it dull/unpolished but yet 
> still look like clean metal. My attempts at micro and macro normals 
> (using variations of the textures that Ive posted in the metals.inc 
> thread a while ago) didn't work well, and there are very fine lines 
> between completely black, way-too-bright, and fuzzy.

Some ideas to try:

- As the bell's texture lives primarily from reflecting the light probe, 
and nothing else does, blurring the probe instead of the texture surface 
could to the trick.

- Add a bit of deformation to the bell, using pertubed normals (some 
subtle bumps), to "break" the overly-ideal look of the bell.

- Use an averaged texture, with the major component as-is and another 
one with significant micro normal pertubations, to simulate tiny 
scratches in an otherwise well-polished surface.

- Maybe add a coat-of-arms or something, etched into the outside of the 
bell.


You should also do something to the belt, to break up the "sterility". 
Some leather-like crackle normals (with an offset value) maybe.

The clapper suffers from not having bevelled edges, maybe a cylindrical 
shaft would do a lot better.


> Render time 2:06:31 on a quad-core 2.6GHz, beta33.

Appears a bit long to me. Then again, I'm not too familiar with 
rendering times of spline-based shapes.

 > Minor focal blur,
> radiosity with decreased brightness, and area light used.

I think it shows that you decreased radiosity brightness: It should be 
stronger. I can't exactly pinpoint why, but maybe it's the way the bell 
and the handle appear to be lit totally differently.

(just my $0.02)


 > I think the
> line up the middle of the inside of the bell is a seam in the HDR probe 
> (which is St.Peter's Basilica from Paul Debevec's gallery), I will 
> rotate it once I have more changes to make.

Given the symmetry of that line, I'm pretty sure it's not in the light 
probe, but the bell reflecting its own rim. Which in turn indicates that 
the reflection is a bit too strong.

If you're using them right, the Devebec light probes are perfectly seamless.

Post a reply to this message

clipka wrote:
> Some ideas to try:
> 
> - As the bell's texture lives primarily from reflecting the light probe, 
> and nothing else does, blurring the probe instead of the texture surface 
> could to the trick.
Possible, but I'd like to make it possible to reuse the model in other 
places, so blurring the background is something I'll leave as a last option.

> - Add a bit of deformation to the bell, using pertubed normals (some 
> subtle bumps), to "break" the overly-ideal look of the bell.
I did that a bit on the inside (it's a slightly different finish/normal 
but the same pigment), but evidently not enough. I am content with the 
outside as-is, because a newly polished bell does look very close to 
that. I'm going to try to add 'rings' on the inside though, as evidence 
of the lathe tool used to shape the metal.

> - Use an averaged texture, with the major component as-is and another 
> one with significant micro normal pertubations, to simulate tiny 
> scratches in an otherwise well-polished surface.
I haven't played with averaged textures yet, seems like a good thing to 
look into.

> - Maybe add a coat-of-arms or something, etched into the outside of the 
> bell.
Not in the plan right now, but who knows.

> You should also do something to the belt, to break up the "sterility". 
> Some leather-like crackle normals (with an offset value) maybe.
It is supposed to be fairly smooth plastic, but it does need something.

> The clapper suffers from not having bevelled edges, maybe a cylindrical 
> shaft would do a lot better.
I'm trying for realism in the model here, and the shapes of everything 
are very close right now to what a real bell looks like. The shaft 
actually is a hexagon, though I could round over the edges of the 
clapper itself a little more.

>  > Minor focal blur,
>> radiosity with decreased brightness, and area light used.
> 
> I think it shows that you decreased radiosity brightness: It should be 
> stronger. I can't exactly pinpoint why, but maybe it's the way the bell 
> and the handle appear to be lit totally differently.
> 
> (just my $0.02)
I found that using rad brightness 1.0 made a huge bright glare on the 
side of the bell, which also suggests that there is too much reflection.
Another reason the lighting is a bit odd is that I added a spotlight a 
ways behind the camera to illuminate the inside, as increasing the 
ambient of the light probe made the whole thing glow. Gotta reduce that 
reflection I see now.

>  > I think the
>> line up the middle of the inside of the bell is a seam in the HDR 
>> probe (which is St.Peter's Basilica from Paul Debevec's gallery), I 
>> will rotate it once I have more changes to make.
> 
> Given the symmetry of that line, I'm pretty sure it's not in the light 
> probe, but the bell reflecting its own rim. Which in turn indicates that 
> the reflection is a bit too strong.
> 
> If you're using them right, the Devebec light probes are perfectly 
> seamless.
Right you are, I rotated the probe and got the same effect. I wasn't 
sure though, since I had to convert it to spherical mapping in HDRShop, 
as b33 doesn't support map_type 7.

Thanks for the input, I'll be working on it more.
For reference, this is the type of bell I'm trying to model: 
http://www.schulmerichbells.com/products_hb_anatomy_of_handbell.php

CShake

Post a reply to this message

CShake schrieb:
> I'm going to try to add 'rings' on the inside though, as evidence 
> of the lathe tool used to shape the metal.

Are you sure these things are produced on a lathe? I'd actually expected 
them to be drawn.

>> You should also do something to the belt, to break up the "sterility". 
>> Some leather-like crackle normals (with an offset value) maybe.
> It is supposed to be fairly smooth plastic, but it does need something.

For plastic it's quite an unusual color.

> I found that using rad brightness 1.0 made a huge bright glare on the 
> side of the bell, which also suggests that there is too much reflection.

Are you using conserve_energy? If not, then you should.

> Another reason the lighting is a bit odd is that I added a spotlight a 
> ways behind the camera to illuminate the inside, as increasing the 
> ambient of the light probe made the whole thing glow.

Define "whole thing". Are you talking about the light probe, or the bell?
> Right you are, I rotated the probe and got the same effect. I wasn't 
> sure though, since I had to convert it to spherical mapping in HDRShop, 
> as b33 doesn't support map_type 7.

Expect HDRShop to make a perfect job, too. I never had any trouble 
(except that I found I needed to flip left/right when using spherical 
mapping with a converted probe instead of map type 7 with the original, 
and I think it also needed some N*90 degrees of rotation).

Post a reply to this message

clipka schrieb:
> Are you sure these things are produced on a lathe? I'd actually expected 
> them to be drawn.

Hm, well, seems like they're typically cast, then finished up on a lathe 
indeed.

Post a reply to this message