Can THEY get LW to do this?

[wgreenlee1]

Wouldnt it be cool...if you could bake volumetrics?
Hypervoxels can do it why hasnt someone divised a post render baked solution to an image plane or something...or is there one?
HD Instance bakes a object and clones it so why not a post render pixel effect for volumetric fog?
But would bakeing take longer than rendering the actual volumetric?

[twidup]

Howard, what settings are you using on teh high res?

[SplineGod]

THere are several other methods for faking volumetrics or speeding it up:
1. Cones made of polysgons with edge transparency and procedural noise.
2. Using sprite mode on volumetric lights and then spinning them to fuzz up the shadowing a bit. If you clone the light and pull it slightly off axis from the other and turn off shadows you can then control the intensity of the volume shadow.
3. Stack and series of polys close together and use procedurals etc to effect the transparency with falloff.
4. Vibrate an object very quickly with motion blur on to create a volume effect. Use textures with world coordinates to noise in the volume.
5. Use Hypervoxels especially HV sprites.

Ok, this is about Volumetrics....or the need for Doublers return?
Im doing a composite with volumetric cones of light....at HD res...and volumetric render times are extremely exponential when you render very hi res!!...a 960x540 render that takes 30 mins takes 9 hours at double the res....now why is that?!

Ive got a background, the lights then some foreground objects.
I did a test in AE...I rendered all 3 elements separately, then comped them.
This saved me considerable time, considering I rendered the volumetrics/lensflares without AA, then comped they look fine...I could of even rendered them half res and then upres'ed them in AE...but then fields become useless...
Now...If I rendered them in LW without AA, they would of looked like crap where an object intersected them, but comping in AE gave me no problems...SO...
why cant LW do this internally?! why the outrageous render times when AA is needed for volumetrics?...cant LW figure a way to render the volumetrics quickly at low res and then apply them like a layer in AE?......why did Doubler dissapear? [/QUOTE]

[lightwolf]

Originally posted by HowardM
Im doing a composite with volumetric cones of light....at HD res...and volumetric render times are extremely exponential when you render very hi res!!...a 960x540 render that takes 30 mins takes 9 hours at double the res....now why is that?!
Hi Howard,
I've done a bit of volumetric coding, so I guess I can answer that question.
At a higher res, the raymarcher (which is the bit of code that steps through the scene for every pixel being rendered, evaluating the volumetrics) uses a smaller steps size -> more steps. This is why it takes longer to render. It has to do that to avoid rendering artifacts.
So, doubling the res means rendering 4 times as many pixels, with roughly 4 times as many steps in the raymarcher, resulting in a factor of 16.
Hm, 30 Minutes x 16 = 8 hours. Close enough. :)
Cheers,
Mike

[HowardM]

Thanks LightWolf for that great explanation!
So, this is definitely a reason why Doubler should return to LW!
I am rendering at 1920x1080 for HD, and we use fields because our flying logos are crisper than anyones! :)....so I NEED to render my volumetrics at full res with fields....and this is why I desperately need Doubler so I can still use fields!

Thank Larry for those techniques, but its funny, why cant LW just do all this itself? I mean really, if we can do MB spinning tricks, why havent programmers written this into LW? Area lights are worthless, but we can workaround by spinning a light...I think its funny that we have to workaround things like that...why not either make Area lights work faster or make an Area light just be a MB spinning light and dont tell anyone!!! :)
Have choices like, Volume, Sprite, or Spinning Trick!....its just funny that we have workarounds for things that should work well in LW to begin with.....does ANYONE actually use real volumetric lights for film?!....HVs are great, but who really uses them for film or anything large?!.....so why in the hell have them!?

..also, I am using HV sprites for another anim that is taking 9 hours a frame compared to 30 mins at half res! SPRITES! cant imagine what Volumes would take!

...so really, why did Doubler dissapear?!...I really could use it!

[lightwolf]

Originally posted by HowardM
...so really, why did Doubler dissapear?!...I really could use it!
Hi again,
well, I guess because volumetrics are integrated in the rendering pipeline now. They used to be a pixel filter, and doubler was, if I remember correctly, a combination of the pixel filter that only rendered every second pixel, and an image filter that interpolated the samples.
Now it is a volumetric handler that is integrated into the renderer, which is why volumetrics cast shadows, reflect and refract.
You could always render at half res and slow down your animation to half speed, and interlace when you comp...

Volumetrics could probably be speeded up as well. I wrote a replacement ground fog, that is around 3 times as fast as the original code (which isn't very mich optimized, and only a sample in the SDK).
I'd like to see the quality settings of the current incarnation replaced with a quality % and a dithering %. Dithering the columetric samples can help a lot in reducing the amount of samples, thus speeding up the process.
cheers,
Mike

[HowardM]

hmmm, yeah I guess I could render half res and half speed and still get away with faster times.....hmmm...I guess these workarounds make a resourceful artist!

thanks again for the details....so why cant they make a Doubler for the new volumetrics?...because it isnt a pixel filter?
hmmm, then yes they need better methods or controls, for hi res renders...Im not even freaking using textures and it still takes forever at LOW settings!
:(

[lightwolf]

Originally posted by HowardM
thanks again for the details....so why cant they make a Doubler for the new volumetrics?...because it isnt a pixel filter?
hmmm, then yes they need better methods or controls, for hi res renders...Im not even freaking using textures and it still takes forever at LOW settings!
:(
You're welcome.
Even with a pixel filter it is tricky (kind of the same problems form a programming perspective that XDof, digital confusion and others have). Pixel filter see an image while it is being rendered, pixel by pixel (which, theoretically could be _any_ pixel, not just one after the other in a line).
Because of this they can't apply any operations that go beyond the current pixel (like getting the value of another pixel and blending it, wat doubler did).
It can be done in combination with an image filter (->xdof, digital confusion), by using a pixel filter that stores the pixels in a temp storage area, and the image filter integrating those temp pixels to create the final image.
Now you could say, why not use an image filter in the first place. That is because you can'T for example shoot rays (i.e. raytrace) into the scene to get certain information.
But agreed, better control would be cool. As I said, I got speed gains of 300% just by using controls that are more tweakeable (I had more dithering in my image then, but it looked good).
Cheers,
Mike

[HowardM]

well come on buddy, where is the 300% faster rewrite of HVs and Vs?!?! :) GET ON IT! :p

[zuzzabuzz]

Perhaps a nice Lightwave/Newtek solution would be letting people adjust the step size?
Before I got Lightwave, I used Imagine. They had finally added volumetrics to it, and gave step size as an option for rendering. It was quite handy.
For certain types of volumes, step size is not that important/noticable. What would be even cooler would be
STEP SIZE ___ [E] with the E button for varying it within an animation.

[edit] -- Hmm..kind of similar to what Lightwolf already suggested above. :-)

[lightwolf]

Originally posted by HowardM
well come on buddy, where is the 300% faster rewrite of HVs and Vs?!?! :) GET ON IT! :p
:shrug: I'd have to re-code everything, and integrating Vs properly would be, bluntly, a royal pain. But I did pass my suggestions on :) I'm not sure how well HVs and Vs are optimized anyhow, it wouldn't probably make them that much faster. Volumetrics are expensive to render. Ground fog is only a very simple case. :hmm:

[lightwolf]

Originally posted by zuzzabuzz
Perhaps a nice Lightwave/Newtek solution would be letting people adjust the step size?
[edit] -- Hmm..kind of similar to what Lightwolf already suggested above. :-)
Yup, it is.
In my code, 100% quality is the optimum step size.
step = optimum step * (1 / quality)
100% = 1.0 quality.
I also added a jitter control in %, so:
step = step + (random * jitter * step)
Which lets you get away with a lower quality, but introduces noise.

[SplineGod]

Originally posted by HowardM


Thank Larry for those techniques, but its funny, why cant LW just do all this itself? I mean really, if we can do MB spinning tricks, why havent programmers written this into LW? Area lights are worthless, but we can workaround by spinning a light...I think its funny that we have to workaround things like that...why not either make Area lights work faster or make an Area light just be a MB spinning light and dont tell anyone!!! :)
Have choices like, Volume, Sprite, or Spinning Trick!....its just funny that we have workarounds for things that should work well in LW to begin with.....does ANYONE actually use real volumetric lights for film?!....HVs are great, but who really uses them for film or anything large?!.....so why in the hell have them!?

..also, I am using HV sprites for another anim that is taking 9 hours a frame compared to 30 mins at half res! SPRITES! cant imagine what Volumes would take!

...so really, why did Doubler dissapear?!...I really could use it!
I agree in principle but n reality IM just glad that LW is not so hard coded that we dont have tricks or other options. I dont look at these things as workarounds but as just a new set of paintbrushes. Im not necessarily looking for an accurate simulation but flexibility in how I can apply the tools :)

[CGTalk Moderation]

This thread has been automatically closed as it remained inactive for 12 months. If you wish to continue the discussion, please create a new thread in the appropriate forum.