Home VRay.info Shop News About Contact Advertise Write for us

Topic: Glass Opacity: How clear is it?

posted by MPGREGG
archived on 30.4.2002

My that furniture looks awfully familiar. Can't beat a classic. This is a WIP of an appartment. I had to use a lot of samples. I do have a few interior lights which help fill in the shadow areas. One thing I had to do was turn off my window glass. Whenever I turned it on my rendering times took a steep nose dive. A test rendering @ 300x300 took 58 seconds w/o the glass and 6m34s w/ glass. I was under a deadline so I had to nuke the glass in the final image. When I have more time I'll try checking my setting to see if I was doing something bad with my glass.

Follow-ups

A: Yes... glass can be quite painful - though you can sometimes render a pic with no window glass, then camera map it onto selfillumed geometry and render a second pass for your reflections/env only with no lights or GI or anything - that way it's pretty damn fast, and you get almost the same result - only problem is that reflecting around corners/away from camera doesn't work at all well...

Comment: What are you using for interp samples? A high number gives very clean GI, but it's blurry as hell - I find that something like 8 gives GI as sharp as possible without actually showing harsh boundaries from the irradiance map - only problem then is your GI samples needs to go up.
For optimally clear GI, try something like 8 interp samples and a high irradiance map GI sample rate until it stops being blotchy - problem is this can mean *huge* sample rates. So maybe go for something like 14, and a somewhat smaller interp samples.
Or use direct computation - 'tis extra slow, but your GI has ideal contrast - no extra blurring at all.
Changing interp samples to a value that doesn't squash all your nice GI detail can turn a scene that looks like it's practically lit with ambient light into one that has really nice contrast and clarity - though the blotches go up...
Bottom line is, you will get a big rendertime hike, but a big quality boost...

Q: What worries us about lowering interp samples and raising the GI sample rates are render times. They can become huge!
A: Yes, that's the good ol' quality/speed tradeoff... careful selection of sampling sensitivity and rates can help a fair bit, but in the end, if you want sharp, clear, extremely high quality renders they will take a lot longer than blurry or blotchy really blazing fast renders...

Q: Maybe some sort of exposure control would do the trick. (cheap, but effective).
A: Yes - if you play around in photoshop you can probably get a very nice result...
What would be really cool would be custom intensity remappings - so you could do things like simulating the light response of film... and handle things like these issues in architectural renders.

Q: I still don’t understand why I have to the Environment override to 3 o 4 in the multiplier spinner to get enough light in. If it didn’t burn out exterior and objects reflecting the environment it would be no problem though.
A: /begin speech standard_speech_42
It's the issue of dynamic range - apart from clarity issues caused by fuzzing from accel techniques, the problems stem from the fact that interiors really are very dim compared to exteriors but we don't notice it because of impressive abilities of the human visual system to show a balanced picture of things with big intensity differences. So the outside really is that bright, it just doesn't look like it.
See, the way the eye perceives things, it has a great deal of detail in darker shades and less for bright ones. Combined with it's freaky automatic light level control in the shape of a constantly adapting pupil size, this allows perception of an extraordinary range of intensities - consider that you cannot see stars during the day, yet at night they are very clear.
Low dynamic ranges are all that is available on a computer monitor. It physically cannot display light brighter than a certain amount. So for displaying things like interiors on it, if you try to display it so that the darker parts are clearly visible and look as bright as they do IRL, the bright parts get blown out.
To get the same *perceptual* effect on a computer screen as you do when looking at things in real life, you need to do intensity balancing to make the dimmer shades brighter (so they have more perceived detail), and the brighter shades compressed (so they have less)... gamma correction is useful for this, and so are more complex custom adjustments.
The thing to remember is that these aren't actually really kludges - while they no longer represent something twice as bright with twice the pixel value, you are working on an artificial display anyway - the computer monitor has colour and intensity reproduction characteristics that are not much like reality in many cases - so if a modified image creates a perceived effect closer to the situation in reality, by all means, make that modification - it makes it *more* realistic from a human point of view, not less.
/end speech

Anyway, see how you go with a somewhat darker skylight and maybe gamma correction with a gamma of 1.8 to start with (assuming you have a display calibrated for linear gamma) - or some fun with levels/curves...
Only problem is that you really need 16-bit processing to avoid banding/noise, and while MAX exports it very nicely in the PNG format, Photoshop has a broken PNG importer than doesn't handle 16bpc PNGs, even though Photoshop can do 16bpc calculations and 16bpc is supported in the PNG spec...

> print this entry

VRay.info v2.0 | VRay.info is NOT the official VRay site. VRay is property of Chaos Group