Here is the error:
depthbuffer_singleface::resample() - Input size: (size2t 1920,1080) is not a valid supersample of: (size2t 512,512)I have a 32 bit full float exr as depthmap.
Cheers.
Hi Juan,
We received your question at 5:30am Pacific Time, I have forwarded it to the lead developer to take a look.
Please keep in mind both me and Radka are now back in Canada and are 8 to 10 hours behind Europe, so there will be some delay in the communication…
All developers are in North America, too.
Cheers,
Bobo
Thanks Bobo.
I know but i forgot! 
I’m sorry, i’m so stressed with this project that I completely forgot the time difference, I’m sorry, I’ll try to post my questions during night hehe.
It seemd odd to me that there was no answer to the other post, and since this problem was more urgent I wrote to everyone, next time i’ll develop more patience, i’m sorry to be such a nerve 
Thanks Bobo.
Cheers.
This appears to be a bug. Currently Krakatoa will only accept a matte object depth map if its size is a multiple of the camera resolution. In this case 1920 is not a multiple of 512 (and 1080 is not either) so Krakatoa is throwing that error. In the mean time, please resize your depth image to a multiple of your camera size.
What do you mean with the camera resolution?
My output render size is 1920x1080, the matte sequence is also at 1920x1080
Cheers.
We were able to reproduce the problem internally. It is actually a bunch of problems within the same feature.
I went back to v2.0.2 and 1.6.x and discovered that the bug was introduced in release 2 and it works right in 1.6.x.
Unfortunately, I suspect 1.6.x would not be enough for you to render your scenes. We have a build called 1.6.2 which uses the same “thinkbox” license as 2.x if you want to try, but if you are using Magma on your flows, you would have to redo them using the old KCM modifier, and some features would be missing.
It would be possible to work around the problem in 2.x, but it is quite convoluted and would require time and disk space.
Basically we have three issues here -
*Initial Depth Maps in 2 require the >Enable Matte Objects to be on even if there aren’t any (in 1.6, the two are independent)
*The resolution of the Initial Depth Map cannot be lower than the resolution of the internal depth map (which is scaled up by the Super Sampling value)
*If Shadow Casting is enabled in any scene lights, the shadow size collides with the Initial Depth Map size (which explains the 512 in the error message).
So it would be possible to render a pre-lit scene without shadow calculations of the SuperSampling is set to 1 or if the Z-Depth matches the SuperSampled depth map size.
Krakatoa has the option to save PRTs with the Lighting baked into the Emission channel in order to render later without re-calculating lighting.
So if you would save your particles to a new PRT sequence with “>Compute Lighting And Copy Lighting Channel Into Emission Channel When Saving” checked, then load that in a PRT Loader, hide everything else, check “>Ignore Scene Lights”, “>Use Emission”, “>Use Matte Objects” and “>Use Depth Map Files” with Super Sampling set to 1, your particles will be occluded by the Z-Depth pass. Obviously, if you have half a billion particles per frame, this resaving step would be out of the question…
We will look into fixing this bug, but it might take a week. 
I think your workaround will work, I can afford the resave time, teh particle simulation is fairly simple, so i think i can rebake and reuse it
Thanks Bobo and Darcy, i’m glad you find the problem.
Cheers!
Please note that you should be able to render a Z-Depth pass at a higher resolution (for example 2x the camera resolution) if you want to use Super Sampling.
Right now Krakatoa will not up-res the Initial Depth Map sequence, but if you provide a 2x, 3x or 4x depth map, it should work.
Optimally, if you want a SuperSampling of 2, you would have to provide a Depth image that is 2x your image output resolution…
What is the reason for using a Depth Map sequence? Are you masking by render-time-displaced geometry or something like that?
My main render engine is iRay, and i’m integrating the particles in a shot where i use Motion Blur and Depth of Field, so the only way i’ve found to do a good integration that uses that is the Depth Map because i can’t make a good mask of a blurred object manually, so i use iRay for the main render, then i use mental ray to extract the mattes and the Z channel with the motion blur and the dof applied and i use that as matte in Krakatoa, it should work, but i have to try it, so far i just did a quick test, and it seems to work.
The Krakatoa DOF seems to do a good job too but with a problem, it don’t blur the matte object edges, so the dof is applied to particles, but not to mattes, so the only way was using the multipass, but matching blurs is hard, so i plan to use the zdepth to use it as matte because of that.
Thanks for the supersampling advice, i’ll try to use a x2 zdepth, should i use the supersampling value and put it in 2 ?
Cheers.
Yes, that will ensure that the internal buffer of Krakatoa is set to 2x the camera output, and the depth map will just load into it without resampling.
Ok, thanks.
I’ve just rendered a sequence at 1080p x2, i’ll see how it goes
Cheers.
Hola Juan,
We have posted a 2.1.5 build of Krakatoa which should have the Initial Depth Map errors fixed:
thinkboxsoftware.com/krakato … ilds-beta/
Here is the Change Log for it:
thinkboxsoftware.com/krakato … change-log
Please let us know whether it works for you or not.
This is still a Public Beta and might have other issues.