Why IrMap over Brute force for animation rendering?

[francomanko]

A case of development overtaking documentation. :P
Working on it.

great to read this but I have to say, the past few months for me has been a nightmare due to really poor documentation ( I don't mean vray, but with other software) I've spent more time searching google for solutions than actual work ( well it feels that way) Documentation should really always be updated as a priority and i dont really want to rely on being lucky and reading the 20th post on a forum thread to find the answer

[^Lele^]

Documentation should really always be updated as a priority

I wholeheartedly agree.

[NicoC]

I'm still puzzeled. What's the point of doing a LC calculation before each frame when 1 flytrough calculation yields good results ?

How much ime does it take to render the LC for each frame on your side ?

I must admitt I agree with Pixelab here. As far as I understand Lele's explanations in previous posts, the LC precalc of one full sequence is less flexible: may you need to change an object visible from frame 150 to 185, that would require a full recalc. So, if I understand this correctly, if you know you won't need to change a thing, LC full precalc would still be the way to go ?

Maybe it also depends on the number of shots you need to produce: maybe for a few ones, handling precalc for each wouldn't be too much bothering, but for many, you may prefer to send everything in one go. Less human time / more CPU time VS more human time / less CPU time, something like that.

Like stated before, I guess we heavily need updated documentation from Chaos Group, along with one or two cases studies

Again, thanks A LOT to Lele for his great explanations. Even if I come here regularly, I just discovered this thread yesterday ... a must.

Cheers.

[^Lele^]

I must admitt I agree with Pixelab here. As far as I understand Lele's explanations in previous posts, the LC precalc of one full sequence is less flexible: may you need to change an object visible from frame 150 to 185, that would require a full recalc. So, if I understand this correctly, if you know you won't need to change a thing, LC full precalc would still be the way to go ?

Well, it's a bit more nuanced than just moving an object.
Because the samples have to be in world space and size, their coverage across a sequence is fully unknown.
It's akin to thinking of spraying a flat with a fixed number of drops of paint, of a fixed size: you have to assume, or test and interpolate, a lot of things.
You will only find out at the very end if your calculations were really correct: you may find you ran out of paint 3 frames from the end, and there may be no obvious way to pick up the painting from a few frames earlier and continue it on without a visible change in the results.
The new approach always takes care of painting correctly what it sees on the single frame, and will do so regardless of contents, and of course sequence length.
To keep to the painting analogy, per-frame, Leak-preventing LC is like painting with a smart can of paint which will automatically spray more droplets of the right size in the places which will need more, all without the user even thinking about it.
You'll never run out of paint, and will always use the right amount for what's to be painted, unerringly so.

[pixelplume]

great analogy

[NicoC]

great analogy

Indeed ! Lele - what would we do without you ?

[^Lele^]

Indeed ! Lele - what would we do without you ?

Save on spat bile?^^

[NicoC]

Save on spat bile?^^

ahahah yes probably too

[dorham]

Hey,

I've just try your settings and it's flickering if I don't precalc my LC :/
It's a VR movie, do you think it can come from Spherical ? BTW congrats for 3.5, same scene is x2 faster to render
Thx

[Morne]

Hey,

I've just try your settings and it's flickering if I don't precalc my LC :/
It's a VR movie, do you think it can come from Spherical ? BTW congrats for 3.5, same scene is x2 faster to render
Thx

Did you split the channels out? It's worth playing the movie per channel. Might not be GI flicker, may be something like AA or specular

[dorham]

I'll give it a try. I've made another test with higher retrace and it's all good
Thx

[Magnum_design]

So why dont just make a slider with quality-settings from draft to ultra and an animation tickbox. Ive been reading and trying diffrent solutions for days before I found this thread.

[^Lele^]

So why dont just make a slider with quality-settings from draft to ultra and an animation tickbox. Ive been reading and trying diffrent solutions for days before I found this thread.

i actually made one.
It had the same exact amount of controls as doing it by hand for what concerned the sampling part.
For the GI, perhaps it can be made a bit more automatic, today, with a bit fewer controls, but not by a lot (ie. one slider instead of two spinners, and at the cost of flexibility, as can be seen above where LC subdivs were correct, and only retracing needed some bumping in the specific scenario. hardly a game changer, possibly confusing, surely not efficient as it'd have to raise both subdivs AND retrace together, leading potentially to rendertime explosions.).

I think the best way to approach this is to have it well explained in the docs. I'll see what can be done.

[rikou]

... I think the best way to approach this is to have it well explained in the docs. I'll see what can be done.

+1 !!

Really need it

[piotrus3333]

very impressing.
Did not believe so I had to render something.
so funny most of us missed that...