I've been doing some tests in P6 using the new IBL and AO features. The above image is a screencap of the video clip for those who can't play movies for some reason. Overall, this was just a quick test to assertain the quality of IBL/AO for animation purposes. Here's a link (or two) to a recent test I ran using the reduced-resolution M3. Here's the links (Quicktime required):

IBL/AO Test - Broadband version (824kb)

IBL/AO Test - Dialup version (338kb)

IBL with AO was used obviously, plus one key light (spot) with shadow map, and one kicker light (point) with no shadows.

Some things I was pleased with...

The AO came out nice and clean, no flicker.
Antialiasing was acceptable; minimal pixel crawling/roping at medium settings, and virtually none at higher settings.

Some things I was not so pleased with...

Rendertime. This clip took approximately 2 min 40 seconds PER frame, which isn't bad until you consider that it's only the one figure with no props or hair. Hair, for instance, would have added a lot more time... a still image of this scene with dynamic hair using AO and the same settings took around 50 minutes at the same size. Trans-mapped prop hair wasn't dramatically better... around 25 minutes. So, in the end, rendering this scene with IBL/AO was much faster than using actual HDRI or GI, and there was no flickering to contend with.

So overall, I was pleased with the results, but the speed still gives me reservations.

Looks awsome Maxxx!!!!

fine example and extrapolates to some decision-making. agree no sign of flicker and I for one can spot it a mile away. I assume you ran at default Poser6 scale with no massive scale-up? Maxx, to me the lighting is quite convincing. This places the capablility of realistic daylight on the Poser worktable, IMO, at a better quality by far and faster than faking with dome lighting. Practical for animation? Your plea for network render is well known, as is my willingness to work with a "RenderCondo" concept. But even then, the frame budgets....wow! I use FinalFantasy/AkiRoss as my rough benchmark...the budget was 90 min/frame at full film/big screen size. I think that if one had to have it for a short film, and if one could throw a handful of CPUs at it, to get this kind of result in Poser...it's pretty damn cool. Now my obvious next question....can you rough this against what you believe your time would be in 3DS Max with MentalRay or whatever you would use to arrive at this quality? What magnitude faster/slower, Max/Poser? I'm not asking you to necessarily go thru it experimentally, just give your seat of the pants opinion. Great work and thanks for the pioneering. In your opinion, was performance steady frame by frame? No slow-down due to the memory leak? ::::: Opera :::::

"Looks awsome Maxxx!!!!" Thanks, ghelmer. I'm not exactly 'proud' of the motion, but a test is a test. I rushed through the keyframing, and it shows. ;-P Operaguy: "I assume you ran at default Poser6 scale with no massive scale-up?" Default scale, no scale-up involved. "can you rough this against what you believe your time would be in 3DS Max with MentalRay or whatever you would use to arrive at this quality? What magnitude faster/slower, Max/Poser?" Good question, and I did do a 'sort of' unorthodox test of this very thing. The same scene, imported into 3dsMax, using Max-specific shaders, etc. I set up a dome light consisting of 91 spotlights to "fake" GI, then added a kicker light and key light, just like I did in Poser. Then I imported the entire PZ3 via a plugin into Max, changed the textures to the Max-equivalent, and rendered out just one single frame as a test. The resulting image only took 45 seconds at the same size output dimensions, and the result looked very similar. So, even with a 91 light dome rig, Max still rendered dramatically faster on a single frame basis. I didn't test Mental Ray, but I'm sure the result would have been about the same. That said, however, Poser still has a major advantage in the way of displacement rendering! If I wanted more detail, for instance, I could have added a displacement map to the texture in Poser, and it wouldn't have increased the rendertime too much. If I wanted sub-poly displacement in Max, it probably would have tripled or quadrupled the rendertime, and probably even more if I needed very high antialiasing too. "In your opinion, was performance steady frame by frame? No slow-down due to the memory leak?" Unfortunately, I wasn't looking for this. I didn't notice any slow-down due to the memory issue on a per-frame basis, but I suppose it's possible. I only watched the first few frames roll off. ;-)

Looks good maxx I haven't been able to get a frame rendered in under 10 minutes using IBL and AO, so I'd say that your times were very good. The only way I've been able to get down to the 10 minutes mark was to use texture filtering. Mine had more figures (V3 and Laura). Did you use texture filtering at all? Ed

"The only way I've been able to get down to the 10 minutes mark was to use texture filtering. Mine had more figures (V3 and Laura).

Did you use the reduced resolution V3? This test clip uses the RR M3, which renders a bit faster than the full res version. Two or more figures with props, etc. in a scene will definitely increase the rendertime, and also any kind of transparency (such as transmapped hair), combined with AO, will slow things down dramatically as well. After all, AO is raytraced.

"Did you use texture filtering at all?"
I never use texture filtering in my renders unless I have a patterned texture like, for instance, a checkerboard or weave pattern to render, which gets "mottled" without it. Although it may improve speed slightly in some cases, and it improves textures in some cases, I find texture filtering seems to interfere with most of my textures, like skin for example. Message edited on: 04/11/2005 04:22

No my figures were a heavily morphed V3 and a morphed and scaled Laura with a pic rendered in Vue as the background and light source. The reason I was using texture filtering was the size of the textures.

All I can say is WOW! I thought it was an image from a "real" film clip.

very nice indeed maxx!

"In your opinion, was performance steady frame by frame? No slow-down due to the memory leak?" Unfortunately, I wasn't looking for this. I didn't notice any slow-down due to the memory issue on a per-frame basis, but I suppose it's possible. I only watched the first few frames roll off. ;-) " ----------------------------------- well, it is possible speed will improve with SR1 becuase there defiately could have been slow-down during the render, even if it did not crash. ::::: Opera :::::

Fantastic work Max. Looks great. This might be an old trick - but for rendering strand (and maybe trans) hair with AO a little quicker, just turn on Visible in Ray Tracing off for the hair props. Obviously you need a depth map shadow light in the scene to get the hair to shadow properly using this method.

Good idea, face_off. I'm going to try that meathod next. It didn't even occur to me to do that. I was thinking more in terms of rendering in multiple "passes", but yours is the better idea. ;-)