PDA

View Full Version : My plea to Newtek for LW10!



dmack
10-22-2008, 07:33 AM
OK, for me, if just the following were done in 10, I'd be really really pleased...All to do with the same area

1. Cleverly interpolated depth of field so we can have really smooth strong dof

2. This dof working in conjunction with volumetrics (so dof works with smoke)

3. Volumetric motion blur working with perspective camera (ie item movement taken into account)


PLEASE newtek, make this a priority for 10! It's the foundations like this that really enable LW to be a workhorse app.

BeeVee
10-22-2008, 07:54 AM
Hey David,

Have you tried photoreal moblur/DoF? Seems to work with HVs for me. Here's an extremely quick test I did (30 secs).

B
PS. I did it with surface HVs to better show the DoF but it also works with Volume HVs.

PPS. Best of all, you can see it in OGL so you can check how your DoF is looking without needing to F9 (LightWiki (http://www.lightwiki.com/Interactive_Depth_of_Field) has the explanation)

dmack
10-22-2008, 08:41 AM
Hi BeeVee,

Thanks, I didn't realise that the perspective DOF worked so well with HV's (including the transparent sprite sort) BUT.......here's a classic situation I find myself in. It's an interesting illustration of the day to day issues I have with Lightwave on commercial work.

I'm doing a visualization down a tube (blood vessel) and the rendertime on this tube is around 20mins with Classic AA, no dof. Now, I want to have both smoke in the tube and I want strong depth of field. So, I can either use X-Dof and classic AA and I get a similar render time of 20mins-ish and get beautiful smooth depth of field BUT X-Dof doesn't blur HV's correctly based on their opacity, so that doesn't work....OR I can use Lightwave's realistic depth of field using the perspective camera.....BUT to get a strong smooth depth of field blur, I need to use such strong Anti-Aliasing that the rendertime become prohibitive (I mean completely undoable in a commercial setting). So....I'm left with no real answer to the problem....what do I do....remove the smoke, which damages my competitiveness. This pretty much sums up why I'm calling for a clever interpolated system of DOF, so you can get good blurs without the HUGE render hit. Heavy blurring doesn't need more AA cycles, it needs blurring!

So, as you can see, although the perspective camera SEEMS like the solution...it really isn't.....yet.

BeeVee
10-22-2008, 08:49 AM
It doesn't need more AA passes, it needs more Moblur passes. By default they are at 1 which means one motion blur pass for every AA pass. If you up that you can bring down the AA passes, likewise you can use Adaptive sampling to improve it too. I often set my AA to 3, AS on and MB passes to 3 as well to get a nice median.

B

dmack
10-22-2008, 09:27 AM
I've done some tests and as I thought the results just prove the point...

Classic Camera, Enhanced Low AA 1min49s (add X-Dof and there's no real hit)

Perspective Camera 3Mblur AA's, 1 normal AA and an AS of 0.05 gives a time of 5mins19s but nowhere near enough smoothness to the DOF

So again same but 3mblur AA's and 2 normal AA's and an AS of 0.02...and the render time has gone to 16mins and it's not finished. Now, the actual render (at correct resolution) rather than the tests take around 24mins to render with the clasic camera and X-Dof....I only have to do some quick maths to work out that it's going to take many many hours PER frame to render.....undoable.

This is the reality I'm faced with. I'd love it to work but it just doesn't for strong DOF blurs.

OK, it's just finished - 19minutes and guess what....not smooth enough.....now with some clever interpolation, it would....oh, no, I've been there haven't I.... :)

Dodgy
10-22-2008, 02:03 PM
Beevee, volumetrics don't get rendered with the new MB/DOF, they are rendered using the classic style multiple copies over time methods, so this is going to be a problem until NT figure a way to get the new MB working with volumetrics.

dmack
10-23-2008, 01:32 AM
In a quick test I did, I got the new DOF working well with HV's (sprite based is the test I did). The issue reamins though that it's not a plausible solution as the AA system of DOF simply doesn't work on more render intensive scenes.

BeeVee - any comment? I understand that for very simple scenes, it works BUT can you see there's an issue for more render intensive scenes where AA'ing the scene to get DOF just isn't feasible?

Chuck
10-23-2008, 10:44 AM
Dmack, please be sure to submit your feature requests to our Feature Request database at lw-features (at) newtek (dot) com. We are, of course, not commenting on our plans for future versions of LightWave 3D at this time, but please rest assured requests are being reviewed and considered for future editions of the product.

It may also be worthwhile to submit a bug report with a scene that demonstrates the issues you've encountered with the rendering speed with DoF and volumetrics, as another possible approach for getting the issues of concern to you addressed.

dmack
10-23-2008, 10:53 AM
Hi Chuck,

OK, firstly thanks for getting back to me, that alone is reasurring. I'll go the feature request route as it's not a bug per se, just that in a commercial environment the render times become far too high when there is an alternative solution, it would seem.

Thanks

PS GREAT work on the SSS shaders, SSS in LW has really come of age now.

dmack
10-23-2008, 11:40 AM
Done...coming through Chuck!