Krakatoa


#761

Ahh, got it now. Thanks. Trying to figure out how to project some stuff onto particles using this method :hmm:


#762

Per Pixel Camera Map is fully supported - even in the viewport!
Add your bitmap texture to the Camera Map Per Pixel map and you should see the colors in the viewport (works in a Standard Material’s Diffuse slot, too!)


#763

http://lotsofparticles.blogspot.com/

In case you find this kind of stuff interesting…


#764

I would hope that if they’re in this thread they’d be somewhat remotely kind of sort of mostly interested in lots of particles… :smiley:


#765

Life, the Universe, and Lot of Particles, with ketchup, of course :slight_smile:


#766

Got it working now, kind of. Just need to figure out how to make that mapped color stick.


#767

The slow way would be to apply them to Pflow particles, save out a single frame to PRT, bring them back into pflow with a PRT loader Birth op, add your motion to it, than save that whole sequence out to PRT.

So you are basically baking the first frame color to the Pflow Particles


#768

Hi Bobo, I have a few questions regarding the use of the Initial Depth Map Sequence feature in the 1.5 demo, but first I would like to ask if it’s possible for you to post a quick step by step procedure on how to use it, that would be really helpful.

Here are the questions:

  1. Why can’t I use any other image file types aside from .EXR?
  2. The old krakatoa had a button to enable using Depth Maps, why was it removed? (or was it placed somewhere else, I can’t seem to find it ATM)
  3. According to an old post in the thread regarding Depth Maps, I should render a z-depth pass (I usually render in .TGA) then use that in “Normalized Z-Depth” mode and set the near and far Zs same as the Z-Depth element settings. If this is correct, I will encounter question #1.
  4. If I was to use EXR, is it enough to just render it out with only the Z-depth as the Extra Channel?
  5. I wanted to test this because I 'm curious to compare what it would look like using Matte Objects vs. Depth Maps.

I have tested some of my questions above, but so far I can’t seem to get decent results.

Thanks for any help,
Jeff


#769

Hi!

  1. Version 1.5.0 had an issue with Depth Maps in Voxel Mode (it was not implemented ;)). 1.5.1 will fix the problem, expected to be released to public any time soon.

  2. Krakatoa supports natively only EXRs. When you render, we pass the image to Max to save into any format, but since
    a) Krakatoa started as a stand-alone renderer,
    b) is on its way to go back to being stand-alone,
    c) EXR is the only reasonable open file format in existence (ILM wrote it, nuff said)
    d) our company does not use anything else anyway aside from DPX plates,
    we did not want to entangle it too much into the sucky Max bitmap subsystems.
    You will also find out that if you are saving a Background pass with Matte Objects, the Foreground image will be written out by Max in the specified format, but the background will be an EXR no matter what. Let’s call it a “known limitation”.

  3. That is unrelated. The Depth Maps button in the old Krakatoa was a special mode to speed up matte objects processing. When it was on, a ray was shot through every pixel of the final output image and the depth value at the hit point with matte geometry was saved in the map. Then the particle’s depth was compared to that map. When the option was off, each particle caused a ray to be shot into the kd-tree to find out whether to occlude or not. BOTH these methods required raytracing and were slow. So slow it was the slowest feature of v1.1.x hands down.
    In 1.5.x, we replaced the raytracing engine with a scanline renderer (mental images would have called it a “fast rasterizer” ;)). Instead of shooting rays into the scene, we just rasterize the polygons of all matte geometry objects into an internal depth map, then go on as before. As result, we get thousands of times faster performance with the same result. In fact, in Beta builds we could even save the output as an EXR for loading as Initial Depth Map (but we have removed that option for now, hope it will reappear in the future).

  4. How you render your depth maps is up to you - Krakatoa supports the Max style depth maps, both kinds of VRay depth maps and the Gelato-style “distance value as floating point per pixel” depth maps (the original implementation, we added the others for the regular Max users because we were the only users of Gelato ;)) You should match the settings of your Z-Depth render element in Krakatoa and render to EXR instead of TGA.

  5. Do NOT save in a Z-Depth channel inside the EXR, save the Z-Depth as RGB to a separate sequence using a Z-Depth Render Element like you would do with a TGA. Really, all you have to change is the output file extension in the Render Element.

  6. The main reason to use Initial Depth Maps is to get render-time displaced geometry you could not produce outside of your other renderer. For example we used it for displaced water from Gelato on “Journey To The Center Of The Earth 3D”, but you could use it with mental ray or V-Ray render-time displacement or other render-time-only geometry. It is a PITA to manage - if Krakatoa can use geometry from the scene (the new version can handle many millions of polygons in no time), it is the better approach. Also, the Matte Objects now allow Super-Sampling (in 1.5.0 it is called Sub-Divisions, but the name will change soon) to improve anti-aliasing, and supports semi-transparent and opacity-mapped objects correctly.
    I would not use Initial Depth Maps unless there is no way to get the geometry in front of the Krakatoa camera.


#770

hi I’m using version 1.1.3 and I’d like to enable deformation motion blur
how I can do that ??
the button is not active and I can’t use it
do I have to enable something to be able to use this option ?

I have issue with matte object and motion blur

my particles are glue on a animated vertex polygon and I use the object as a matte
but there is no motion blur on the object

if deformation motion blur doesn’t work is it an other way to do it

thx


#771
Deformation Motion Blur is not implemented in 1.1.3. The button was a place holder and was also a way to remind the feature developer to do the job for 1.5.0 ;)
If we would have implemented the deformation motion blur in 1.1.3, it would have been very expensive since the kd-tree would require a rebuilding on each sample. In 1.5.0, the feature is enabled, and since Matte Objects processing is orders of magnitude faster there, it does not affect performance much. 

A quick benchmark:
*10M particles lit by an Omni light render in 22.078 seconds on my machine.
*Same 10M particles occluded by 20 Geospheres with 200 segments / 800K faces each, or 16M faces total, rendered in 28.11 seconds.
*This means that Krakatoa 1.5.0 processed 16M matte faces in 6 seconds.
*In 1.1.3, the same scene rendered in about the same time without matte objects, but when I enabled them, it needed 406.703 seconds…

A workaround in 1.1.3 (as explained in the documentation) is to use the 3ds Max Multipass Motion Blur. It will be even slower because all particles will have to be reloaded and relit for each pass, but it will respect deformations of matte objects.

The real solution is switching to 1.5.0, but I can understand if that is not an option.

#772

Autodesk has published the Prime Focus demos from Siggraph (Paramount did not allow the live broadcasting of the whole presentation because the movie was still not out at that time). This is the presentation from Day 2:

http://area.autodesk.com/player/loader.swf?p=/player/main.swf&f=http://areadownloads.autodesk.com/oc/ibc09/sig09_d2_frantic01.flv


#773

really amazing to see the initial test you guys and how you push around your software! amazing work –


#774

absolutely jawdropping bobo! I love the varied particle tests that you guys did! totally amazing. :buttrock:

As always, thanks for the very detailed reply regarding the Depth Maps! Got it working like a charm. (and as you’ve pointed out, all I needed to do was render the element and save it as an EXR!)

I only have a few suggestions regarding that:
[ol]
[li]Can there be an automatic way to resize+filter the zdepth exr image to fit a final render size?[/li]- The idea is to maybe render a half size zdepth exr pass and use that for the depth map sequence. The reason for the suggestion is that I just jumped in to an almost finished project, and the plates that I got doesn’t have zdepth passes because thier output is NPR. It would just save me some time to be able to render the zdepth passes half the original movie size.
[li]Use a single exr image as a depth map.[/li]- Right now it gives me a “Matte depth map does not exist:” error.

  • Probably useful for static camera scenes, where we won’t have to render a few hundred frames of the same image.
    [/ol]Regarding 1.5.1 - I’ll try and ask the company if they bought the Support and Upgrade… if so, it possible to get a version with the working depth maps for voxel rendering?

#775

Totally great demo! The eating effects were marvelous! I have two questions:

  1. In the collapsible hood shot, how did you do the metal collapsing? Was it cloth and Krakatoa skin wrap? If yes, How did you drive the cloth?
  2. These cool intelligent particle motions, can you elaborate on how did you drive these? That would be cool :slight_smile:

Thanks!


#776

I have a problem when rendering a cache with krakatoa 1.1.3
I have this message each frame during the rendering

– unknow property: “refresh_GUI” in undefined

does anybody know why ??


#777

That could be a bug :slight_smile:
Can you provide more info like Max version (8,9,2008,2009,2010), 32/64 bit?
If 2009 or 2010, do you have VFB Extensions turned on?

Also, please provide a step by step description of what you are doing - “Rendering a cache” cam mean many things. Are you rendering a PRT Loader with particles cached to disk? Or rendering with the PCache enabled? Or rendering with a PFlow Cache? Or with the Box#1 Disk Cache? Please be as specific as possible.

Finally, do you have the Krakatoa GUI open when rendering, or are you just hitting the Render button in the Max UI without opening the Krakatoa floater?


#778

You are right, the Depth Maps handling is a bit underdeveloped, it has not changed since before I started working on Krakatoa… We will look into supporting a single frame and will talk to the developer about the possibility to resize the image (we do this with the Sub-Division spinner in 1.5.0 but the other way round - we render a larger image for antialiasing)

The 1.5.1 version should be out really soon anyway, but yes, anyone with Beta access can download it.


#779

Please remember that this was a team effort. A lot of these tests were done by other people, esp. Laszlo, so credit where credit is due. Also, as mentioned in the presentation, some of the effects were simulated using Level Set tools (which are based on Flood technology and we developed further during the Dragonball production). Of course, Krakatoa 1.5 now uses the same Level Set technology to fill volumes with particles, but when you look at some of the volumetric tests like the cannon or the hood, they are a mesh representation of a level set of the original mesh, eaten away by a procedural map (Laszlo wrote a procedural map that could be controlled precisely to grow where desired). If you have seen Dieter Morgenroth’s (DiMo) level set tests in VRay, this is a similar approach, but is not just render-time geometry. Thus, it can be driven by other systems.

In fact, RnD for the Dragonball and G.I.Joe projects started the same week of 2008, and we discovered that they called for similar technological approaches, just in opposite temporal directions - in DB:E we had to grow stuff, in G.I.Joe we had to destroy it. If you would play the one backwards, you would get the other effect. Of course, in the end they turned out very different, but the initial research was shared and some tools ended up being used for both at one stage or another.

Laszlo used Box #3 for the intelligent motion of particles. I will see if he would like to comment on that.


#780

Thanks! I got the basic idea about the levelsets and the procedural map. And yeah it would be cool if Laszlo comments on his box3 magic :slight_smile: