PDA

View Full Version : animation and GI Brute or interpolated



vonpietro
06-20-2018, 10:28 AM
quick question

in the new lightwave 2018 - is it better to render brute force with no interpolated for an animation?
or is it better to go with interpolated settings

the goal being the least noise.

MSherak
06-20-2018, 10:37 AM
quick question

in the new lightwave 2018 - is it better to render brute force with no interpolated for an animation?
or is it better to go with interpolated settings

the goal being the least noise.

Try interp with the new noise filtering options under the render tab.

Gungho3D
06-20-2018, 06:40 PM
is it better to render brute force with no interpolated for an animation?
or is it better to go with interpolated settings
the goal being the least noise.

There is also the "Interpolated Cache" path, with which you can cache the GI calculations and then rely on that to be re-used across a series of frames - in other words, GI is not getting re-calculated frame by frame because it's already done.

It is ages since I used this as a solution (arch vis in LW 11.6), plus I've never put it to the test in LW2018, so the fine details and settings I can't recall.

What I do remember is that once I played around with settings for while and got the hang of it, stable good-looking end results were able to be tweaked and rendered much faster.

Hope that is a help ...

Axis3d
06-21-2018, 07:39 AM
Since switching to LW 2018, I've been using the Brute Force method with 1 or 2 Rays. I don't have to worry about blotchies or flickering from radiosity anymore. I used to spend so much time trying to minimize that stuff.

Ztreem
06-21-2018, 08:45 AM
I almost always use brute force even though I use LW 2015. Instead of spending time trying to find the right values as it almost never work for animation, I think it's better to spend the time on actual work and let the render farm do the work in brute force.

kolby
06-21-2018, 11:21 AM
I almost always use brute force even though I use LW 2015. Instead of spending time trying to find the right values as it almost never work for animation, I think it's better to spend the time on actual work and let the render farm do the work in brute force.

This is absolutely true until you need 3000 frames of interior architectural visualization and you don't have your own renderfarm because your render times could jump from 30 min/frame to 8 hours/frame.

Now let's look at very cheap Garagefarm (low, medium and high priority):
3000f 30 min/f - $412, $833, $1,658
3000f 8 hours/f - $6,598, $13,340, $26,537
Not to mention that if you choose low priority to save some money, the BF method will take at least 60 days to render v.s. 4 days in interpolated.

So, in some cases, I think it is way better to spent few days to tune the interpolated mode, than unnecessarily spend 6, 12 or even 25k bucks.

samurai_x
06-22-2018, 04:09 AM
This is absolutely true until you need 3000 frames of interior architectural visualization and you don't have your own renderfarm because your render times could jump from 30 min/frame to 8 hours/frame.

Now let's look at very cheap Garagefarm (low, medium and high priority):
3000f 30 min/f - $412, $833, $1,658
3000f 8 hours/f - $6,598, $13,340, $26,537
Not to mention that if you choose low priority to save some money, the BF method will take at least 60 days to render v.s. 4 days in interpolated.

So, in some cases, I think it is way better to spent few days to tune the interpolated mode, than unnecessarily spend 6, 12 or even 25k bucks.

For interior shots both 2015 and 2018 are slow. 2018 even more. I can't imagine using bruteforce for interiors there. Increasing reflection samples is just way too slow compared to the cheats with adaptive sampling in 2015 that can "fix" render issues.

Best to bake the background and render as separate pass. Then use bruteforce for any animated or deforming objects and then composite the result.

erikals
06-22-2018, 04:53 AM
from my tests...

for exterior BF looks to be good
for archviz, interpolated might be better

pixym
06-22-2018, 02:10 PM
For interior, it is better to use some tricks like area lights (not casting shadows) at the windows. You can use interpollated Radiosity if you have a static scene.

erikals
06-22-2018, 02:35 PM
yes, an image, versus an animation, changes the game.

jwiede
06-22-2018, 03:12 PM
yes, an image, versus an animation, changes the game.

Fine, but realistically, these issues highlight significant deficiencies in the new engine. Having to engage in workarounds like swapping environments for shadowless area lights, or playing all sorts of games to avoid sun-created fireflies in outdoors scenes results in time cost as well as quality costs in most cases.

The real answer is that interpolated needs to be stabilized so it's more functional/reliable/predictable in behavior as it was in LW2015 (right now, there's a hair-fine line between "tolerable" and "unacceptably blotchy" w.r.t. interpolated GI). Likewise, something needs to be done to address the general penchant for indirect specular hotspots/fireflies in the engine. As is, it's possible to "get by" at cost of quality and added tuning time, but it's unreasonable to expect customers to continue accepting that workaround indefinitely.

Dan Ritchie
06-22-2018, 04:20 PM
I tend to use brute force. It takes way too many samples to reduce flashing blotches in interpolated animations, where noise is more natural looking and I can use less samples.

erikals
06-22-2018, 05:15 PM
http://forums.newtek.com/images/misc/quote_icon.pngjwiede  The real answer is that interpolated needs to be stabilized so it's more functional/reliable/predictable in behavior as it was in LW2015
yes, might be. 2018 seemed a little "advanced / time-consuming"


http://forums.newtek.com/images/misc/quote_icon.pngDan Ritchie  ...It takes way too many samples to reduce flashing blotches in interpolated animations
yes, in almost all cases, agree.
it's also better to use a temporal video denoiser on fine noise, compared to tricky splotches.

jbrookes
06-26-2018, 10:43 AM
Fine, but realistically, these issues highlight significant deficiencies in the new engine. Having to engage in workarounds like swapping environments for shadowless area lights, or playing all sorts of games to avoid sun-created fireflies in outdoors scenes results in time cost as well as quality costs in most cases.

The real answer is that interpolated needs to be stabilized so it's more functional/reliable/predictable in behavior as it was in LW2015 (right now, there's a hair-fine line between "tolerable" and "unacceptably blotchy" w.r.t. interpolated GI). Likewise, something needs to be done to address the general penchant for indirect specular hotspots/fireflies in the engine. As is, it's possible to "get by" at cost of quality and added tuning time, but it's unreasonable to expect customers to continue accepting that workaround indefinitely.

Agreed.

RPSchmidt
06-26-2018, 02:16 PM
If you are trying to get the least amount of noise, then brute force is the way to go. Just be ready for the time sink.