PDA

View Full Version : V-Ray Irradiance Map Settings for Animation


Lestat
03-08-2007, 09:38 PM
I've searched, but haven't been able to find anything conclusive. When I render out an animation in V-Ray I get flickering, which I assume is because the Irradiance map changes a little every frame. Is there anyway to prevent this?

The animation has a static camera, and the model "whooshes" in from the side and rotates a bit, so most of the scene is static.

I tried Multiframe Add, but the prepass calcs ended up taking really, really long 50-60 frames in, compared to single frame, probably because the Irradiance map became hugely detailed.

Any suggestions?

Link to a really rough version of the animation:
http://wireframed.com/download/sony_dsc-t10_ani_04.wmv

L.

JohnnyRandom
03-09-2007, 12:34 AM
Bake the irradiance map. IE check "Don't render final image" global switches, switch to "multiframe incremental" and render out as many irradiance maps angles as you need to get you desired level of detail. Check the map in the irradiance map viewer to see the result, add more angles if necessary. You can do this from frame 1 in a perspective viewport if you want. Literally arc rotate around the object and render out single frames.

Then switch too "load from file" load the multiframe map and render. That should cut down on your render times considerably.

Unless of course this what you have already done but it sounds like you are calculating a new map for every frame and adding it to the current irradience map, which could definitely get huge depending on the length of your animation.

Edit: just watched the clip, nicely done BTW, you may want to consider the above but run through the frames, render out the irradiance map from frame 0 to ~ and then check the solution in the viewer see where you may need more samples then add to the map.

-John

MikeBracken
03-09-2007, 02:41 AM
I would suggest using qmc for primary bounces and light cache for secondary bounces. When you have moving objects, you pretty much have to calculate GI for every frame. I have found
that Irmap doesnt usually cut it for moving objects. Also, always make sure you dont have any
interpolated materials.

Regards,
Mike

JohnnyRandom
03-09-2007, 07:37 AM
You already have your irradiance map calculated right? why not just find the frames that flicker and bake in some more samples, instead of recalcutaing all of your GI over again and running the risk of it flickering with a different solution? (it happens) I would at least try and use what you have first before trying to calculate a whole new solution. Reloading your IR map and rendering can take a 10nth of the time or less to rerender the whole sequence.

Since most of the flicker occurs when the object is stationary or near stationary you will not have to calculate every problem frame +-1, for instance at 8 sec to the end the object doesn't even move, and that is where a good portion of the flicker occurs, add some more samples there to your current map with one or two more IR map renders, problem fixed for the last five seconds.:shrug:


Curious, Mike what solution did you use to render your bedroom animation?

Lestat
03-09-2007, 07:59 AM
Thanks guys, I'll take a look at your suggestions. You know how it is with these things, you never have enough time to experiment with the problems that arise. :p

I think Mike may have hit upon the solution - I'm using Adaptive Subdivision on this render, because it was faster than aQMC, and I needed a fast preview.

Do you guys think aQMC will make it go away? Because aQMC gives me the best detail, so I was planning on using that for the final render anyway. I'd really rather avoid having to render out more passes than I need and overcomplicate things, because I'm still relatively new to Vray and animation, and as I said, the deadline looms as ever.

Does using aQMC mean the irradiance map isn't used?

:edit: Oh, and Johnny, I don't have the irradiance map - the irradiance map in the current animation is just a low detail version, the final version will have higher samples and higher resolution.

L.

MikeBracken
03-09-2007, 03:55 PM
Basicaly, the Qmc image sampler is completely seperate from the qmc GI method, which is a brute force method of calculating the GI. What I was suggesting was to use QMC GI for the
first GI bounce and LC for the second GI bounce. As far as the image sampler goes, just use
whatever method works best in this project.

Regards,
Mike

JohnnyRandom
03-09-2007, 08:39 PM
You can get it to go away by increasing the max samples in adaptive subdivision. bump it up one, if it is currently set a 1 you will get 4 samples per pixel, at 2 you get 8 samples per pixel, ect.

As for your GI:

It doesn't matter if the Irradiance map is a "low detail" version you can add to it without issue, there are just more of them with higher settings.

Seriously though, the scene is simple, you have one moving object, and as far as I can tell only one light source other than the GI, the only things reflective in the scene are the camera lens and the lcd screen on the back of the camera, one GI bounce should be more than adequate for what you are doing. Anything else is just well more money out of your pocket waiting, is this ad for HD?

Just trying to save you some time, using a completely different method not a real good idea when a deadline is over your head. You already know what it is going to look like using an irradiance map, that will change if you switch to a different GI method. As I already mentioned you may get flickering using QMC with a "light cache" secondary bounce. Not to mention getting your settings dialed...

I am all for using different methods for GI calcs, but the thing is to use the right method for the job.

Sorry I just think baking Imap is way faster than other solutions for this type of circumstance. anyway cheers and best of luck:)

Lestat
03-10-2007, 06:31 PM
Well, fortunately the deadline has been moved back a couple weeks because they couldn't agree on the script, so I have a bit more time to experiment, which is nice because this is the first time I've been involved with doing any kind of animated content and I want it to look really good even though it's a really small project. :)

Johnny: I completely agree, I don't want to have to tweak lighting and so on all over for a new GI solution, I'd rather just make this one work. When you talk about bounces you're talking about secondary bonces for the QMC solution, right? I've tried setting it to 1 from a default of 3. Doesn't seem to make that huge a difference, but over 350 frames or so, ever little bit helps. :)

I'm still getting some flickering of the background, as if the irradiance map isn't accurate enough. It seems adding more pre-passes doesn't significantly impact the time it takes to render (as it is able to do the higher-res passes faster with more pre-passes, so it seems to almost even out), would I increase the accuracy by increasing these?

Here's the latest video with better settings. I've had to up the samples in the aQMC to a rather obscene (to me) 1/8, because the brushed metal won't come out well if they're lower.

http://wireframed.com/download/sony_dsc-t10_ani_05.avi

(Btw, WMV9 will really cut that size down, pretty impressive).

The rendertime is approx. 2:27/frame for the above quality. (The frames that have the camera fully in the scene). Does this sound reasonable? I'd like to think I could cut it to 1:30 without losing quality?

Oh, and thanks guys for all the help so far, I'm getting valuable info on Vray so I definitely appreciate it. :)

L.

JohnnyRandom
03-10-2007, 07:00 PM
I'm still getting some flickering of the background, as if the irradiance map isn't accurate enough.
L.

Raise the hemispheric subdivisions "HSph. subdivs" (it's under your Irradiance map settings). This will help smooth out blotchiness on your BG plane. Bump it up 10 or 20.

You can also raise the Interpolation Samples as well, this blurs the GI giving it a smoother look,so, I would try raising the HSph Suds, see how it looks, (just render out 30-60 frames or so to get a good idea) and i then if necessary raise the Interpolation Samples. It's all a balance of detail versus time.

Render is starting to look really good :)

EDIT:Render times, well they are all dependent upon the machine you are using to render. 2:27 sounds reasonable for an average workstation of average age (18 to 36 months old). What are you rendering on. You are loading your Imap from file aren't you?

You could cut your rendertime down by turning off secondary bounces, I could be wrong but I don't really think you need them.

CGTalk Moderation
03-10-2007, 07:00 PM
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.