Sonntag, 28. April 2013

Getting back into modeling

I've been re-starting my modeling work for my next project. Loads of head studies using the various references on the web.

I'm using box modeling as my tool of choice at the moment, to get back on track with topology. I will try different approaches later.

This is my head study F - I did some interesting things with the loops flowing back into the characters skull top. Don't know if that works for animation.

Stopped after more or less 3 hours. (trying to get more done in that time)


Montag, 9. April 2012

Two quick notes on using #c4d, #mograph, dynamics, xref and #vray4c4d

While working on Scene B i've found:
  • Xref - even if you activate all the options to modify Parameters for the imported object: you cannot cache simulation onto an imported geometry Xref object. After every "Save" command the baked cache is gone. So I parented the xref under a mograph cloner/fracture object created inside the actual scene. Then I set the null to the same point (weird stuff happens in complicated simulations), after which the baking seems to work.
  • Vray motion blur tag will not work with cloned objects. I had to use the method by Matthias Zabiegly , which worked nicely. 

Sonntag, 8. April 2012

First scene testrender finished in #vray #c4d

I've managed to finish a test render of the opening shot, which is over 20 seconds long - and I'm quite happy that this time it rendered through without any technical difficulties. The technical problems were mainly caused by Apple by switching to a new proprietary implementation of SMB, which basically resulted in a LOT of issues reading files from my server. Here, some general notes on the progress here:
  • Switching to NFS pretty much doubled the speed of the network. There were some problems with permissions, but it's looking good now. Unfortunately, I also need a SMB server on my Lion machine, which Apple changed, ie broke in this release, so I'm looking at alternatives (SMBup looks good). I've spent approx 4-5 hours on the Linux updates alone. Ubuntu is easy to set up, though.
  • Xref When I worked on the textures for the main character, I noticed that C4d doesn't update textures of references every time. I am not quite sure why.
  • The physical camera tag is pretty much unusable for rendering animations. As I am working to scale, the viewport and the rendered picture differ by much more than just a couple of pixels. Maybe I should construct some kind of workaround, where the render camera with the physical camera tag is attached to a normal pinhole camera. At least then I could layout animations.
Here is the first test.


7&UP First Test Scene from Yogi Parish / www.parish.ch on Vimeo.

Mittwoch, 4. April 2012

Working on the first scene of #SevenAndUp in #c4d

I've finally gotten around to more or less finalize the first shot of the film. It is quite long with 500 frames and moved everything over to Vray and Xrefs. Things to note:
  • Scene Scale: I didn't think it a big issue when I started, but I've realized that this is something that I should have payed way more attention to. The exact sizes should be adhered to. The "Scale Document" function has proved invaluable in this regard. Unfortunately, the animation doesn't scale and has to be adjusted manually.
  • Adjusting animation scale. I've found a way to scale it manually in the fcurve window
    • Select the Object Mode to scale - only the object axes are scaled then. I couldn't find a numerical solution to that yet.
    • Pull the top dotted line (indicating the highest value) to the new maximum
  •  Lo-Res Textures for pre-final stage - although I've used generic textures where possible, the size (200MB) slowed down the test renders considerably. I've created rasterized and much smaller versions, which resulted in a speedup of more than 20 seconds for a test render.
  • Lighting is part of the xrefed model - I am not sure that this is a great idea, but works for now.

Sonntag, 1. April 2012

Porting a pool table model to #vray for #c4d

Finally got to work on the pool table model and ported it to Vray for rendering. DOF is very fast. I'm also moving everything to XRefs now, as the model still needs some work here or there.

A good idea might be to use different models in the XRef files - one for the simulation and one for the rendering - therefore switching off the unused one in the Master scene.

Notes on moving everything to Vray and Xref file:
  • Automatic shader translation is good for a quick start, but the shaders have to be re-done anyway, to accomodate to the Vray requirements in terms of look
  • Lighting is quite an issue, but more so on the side of C4D. The viewport will not display the brightness of a light correctly, once we don't use the C4D standard renderer anymore. This is extremely annoying.
  • I've used IES data for the first time - this is quite nice for very smooth gradients from light to dark and much better than the spot lights. Unfortunately, only a few of the studio light makers provide IES. Kinoflo has them, though - which is very helpful. Arri offers a different format (ltd). Unfortunately, the WhitHawkIES library that's downloadable from c4dcafe.com doesn't work with Vray (most of the lights simply come out black)
  • DOF renders surprisingly fast, especially when using the physical camera.
  • If the Vray physical camera gets too close, the viewport and the actual rendering start to differ. In that case, it makes sense to use the Zoom factor on the Vray phyiscal camera tag. This will be a big pain when animating the camera - it would be wonderful to have some kind of automatic correction for the Vray camera. The c4d viewport camera may not be perfect, but for image composition to always fiddle with the Zoom factor, will be a pain.
 I've made some substantial changes to the model, although they are barely visible. (New method for modeling the actual table)


Samstag, 31. März 2012

New version of #AfterEffects GlobalLayerSwitch script 0.9.1 ready.

 Global Layer Switch (Rendering multiple versions)

The new version of Global Layer Switch 0.91 has the following changes:
  • Select the comp from a drop-down menu, but will check if comps are available at all
  • Choose from the list the output template and module.
  • Smaller changes inside the module.
 The new version of Global Layer Switch is available here.

It's free, but if you use it please leave a short comment here, so I know how it's doing. Thx.

Dienstag, 20. März 2012

Overlaying a grid in #C4d camera view ultraquicktip

A quick fix for overlaying a grid on to the perspective view:

I've had the problem when I've wanted to adjust objects in the perspective view approximately to a grid and HUD doesn't offer a grid overlay (or does it?) There are more elaborate solutions out there, but this super-quick fix did it for me.

Simply create a camera deformer and voila: you can add and remove grid lines if you want. Only caveat: To view the grid you need to select the deformer. But for a <10s fix I guess this works alright.

Mittwoch, 29. Februar 2012

#AfterEffects rendering multiple tag versions

I often have to produce several different versions of a composition (usually in different languages) - and, boy, is it easy to make a mistake. I finally sat down and wrote a little script that will allow me to run multiple render versions with one go. It's not perfect but it's a start and it works nicely with Zorro.

Basically, the script parses the comment tag of all layers in all compositions in your project. Depending on your settings in the "On" or "Off" window, it will activate or de-activate a layer including sound. It will then add the extension (3 chars max) to the file name and render out that particular version.

Don't know if this is for use to anyone else, but I'm putting up a link here. Enjoy. Please leave a comment if you think this is useful. (only tested in cs5 but should work with anything after cs3)

Download JSX File

Montag, 2. Januar 2012

#Vray #c4d getting a clean alpha from vraysky

Here is a simple procedure to pull a clean alpha from an object against the Vraysky background. I've had so many white fringes around my object I was foaming, but the setup is simple once you know how it works.
  1. Add a vraysky (VrayLight Tag under Sun/PhysicalSky switch to on) 
  2. In the Sun light Tab switch the Override BG Env to Off, making the sky invisible to the camera.
  3. The camera needs to have a VrayPhysicalCamera tag, otherwise it will not work.
  4. The alpha *has NOT* to set to Straight Alpha in the Save tab. (it is always straight, I messed this up in the last edit, sorry)
Haven't checked out how it works with Multipass On/Off - I had multipass on in my setup.

Sonntag, 1. Januar 2012

#Vray #C4d Antialiasing optimizing - magic DMC settings.

One of the biggest issues for me in Vray4c4d that the forum posts are confusing to me. The documentation is at places hard to understand and I have huge problems understanding the instructions given.

Here I am trying to figure out what was meant in one of the forums for "speeding up animation".

The original tip went like this:

9) a very speedy very high quality render method is to use an universal DMC approach: basically it works in combining high AA subdivision with very high material and light subdivision. for that set the AA to some between 1x32 (or 1x50-1x100), as you like, and set the material and light subdivisions to the maximum value (256 in 1.2.6, in next versions 400). the trick is that the AA will divide the material and light subdivision so if you use the 256 subdivision which looks very high it actually gets divided by the AA subdivision via DMC which is therefore 256 / 32 = 8 the lowest default again. the only difference to default setting is then that you give the very adaptive vray DMC sampler all possible room to decide itself on each pixel how much to subdivide and or antialias. the result are amazing fast and clean renderings:-) 

Ha! Just what I need! Fast and clean renderings. But...!? Where the hell are all those buttons and what do they mean?

So let me try to disentangle this 
  1. I am using Vray4C4d 1.2.6 with R13 Studio. 
  2. Claim: By setting Antialias settings very high in different places we can get speedy renderings.
  3. Some of those settings can be set in the Render settings for Vray.
  4. There are 3 mechanisms for dealing with antialiasing in Vray
    1. Fixed - you know what's coming at you
    2. Adaptive Subdivision - The classic subdivision: The corners of every pixel are evaluated and if their difference is above a threshold, the pixel is subdivided until "Max rate" is reached or the new pixels all have similar colors.
    3. DMC - this is the new (and much smarter) AA method used in Vray. It compares all surrounding pixels and then decides if it should subdivide the pixel. 
Ok, so it's all about the DMC. Two great pointers to read about on the DMC sampler and its inner workings are here:
  • Official manual: http://www.spot3d.com/vray/help/150SP1/render_params_imagesampler.htm
  • And how it really works:  http://www.interstation3d.com/tutorials/vray_dmc_sampler/demistyfing_dmc.html
The essence is that DMC implements a much better mechanism deciding if a pixel should be subdivided for additional rendering compared to adaptive subdivision.

Which means, we have to LET the DMC algorithm decide on its own. The penalty for setting high subdivision max values is small and the image quality WAY better.

Another important point is that all the calculations and the number of subdivisions are related. So lets assume the following example:
  • Material X has Glossiness subdivs of 256
  • The Adaptive DMC image sampler has a Max subdivision of 32
  • Now the actual subdivs are 256/32=8 - because DMC knows that it can subdivide the pixel if it needs to.
  • 8 subdivs are the minimal number of samples that DMC will look at. There is probably no need to go to a lower number. 
The same argument works for lights (shadow subdivision), DOF and Motion Blur.

They all play together, but not quite sure about how right now.

Samstag, 31. Dezember 2011

Creating Grass with Mograph and Vray

Here is the development of the grass patch (seen from a rather great distance) inside Vray. The rendering times had to stay below 5-7 minutes per frame, using GI.

(click on image to see the WIP animation)

Here are some remarks:
  • Visibility tag for distribution of the grass to set to boxes. I needed a LARGE number of instances for this sphere (approx 15000)
  • General orientation of the grass should be upwards - I used a target effector but weighted it lower than 100% to get direction - as last effector in the list.
  • 2-sided materials in Vray made the render considerably slower (factor approx 2-3) and at this distance translucence doesnt matter much.
  • Other cloners to redistribute variations and single plants.
Next step would be SurfaceSPREAD plugin, but I didn't have it at this time.

Sonntag, 18. Dezember 2011

The visibility tag and mograph

For scenes with a very large number of cloned objects I have used the visibility tag, to specifically control the number of clones visible on screen, mainly because my graphics card is rather old.

Unfortunately, the detail in visibility can not be switched off for rendering. This means, that it has to be switched off manually before rendering.

In combination with the "Render Instances" option in the Cloner object the behavior is slightly harder to predict.

Montag, 12. Dezember 2011

Thoughts about motion blur

Below is a shot that shows all the potential problems I can have with motion blur. This is a parallel camera motion to the movement of the white ball, which itself is moving at fast speed across the table.

This particular frame was rendered with the physical renderer. Unfortunately, there is no postpro control on the motion blur, and the frame took 30 minutes to render - highly impractical.

Watch the area, where the two balls overlap - I think that the physical renderer doesn't get it right, but I'm not 100% sure, how the motion blur phenomenon would work on a real cam in this case.


So, we have at least three different objects moving at different speeds
  • the white ball (strong movement)
  • the other balls (motion blur through camera)
  • the table texture with different motion blurs depending on distance from camera.
So I am left with the following solutions to get a good motion blur:
  • High importance: physical renderer and go to a render farm
    • Pro: perfect(?) motion blur, physical accuracy
    • Contra: Problem with setting up the other passes (more research needed)
  • Low importance: plugin in AE 
    • Pro: adjustable in post, might catch the above problem
    • Contra: Doesn't look perfect, complicated object buffer handling, more overhead



Montag, 14. November 2011

Concatenating movies

After a long search I've come across QTCoffee link - a very useful collection of command line tools.
You can easily combine several quicktime movies to a single one, especially for technical trailers and black at the end this tool comes incredibly handy.

Many of the digital playouts I have to produce need different setups.

Here is an example:

catmovie
Bars.mov Black.mov Content.mov Black.mov
-self-contained -force-same-tracks
-o Outputmovie.mov

First tests seem very good - I hope I can use it more.

First caveat: Although concatenated movies of different formats will play back in QT player, after another conversion step the audio was garbled. I therefore use ProRes only before concatenating.


Freitag, 21. Oktober 2011

HDR conversion inside C4D

I have had some ooooold HDR probes sitting somewhere on my harddrive - one day I would buy HDR Studio and convert them...

But actually C4D can convert probes and crosses (HC and VC) in the image viewer to the C4D Longitude/Latitude format for skies. It does it automatically, you open the file and - le voila. A great little gem hidden away!

Sonntag, 16. Oktober 2011

Sorting multi-pass files into directories

The problem of the multi-layered file was that EXtractoR takes forever to extrude eg the motion vector channel, even if the file is on the local hard drive (possibly due to the unzipping) -

the problem is also described here:
http://area.autodesk.com/forum/autodesk-3ds-max/lighting---rendering/exr-compression-and-ae/

I will go for the recommended route and split up the files again. It might also help me with the previews. But we will see for that.


Found the AE Plugin "Immigration" here:
http://aescripts.com/immigration/
for importing files - I would like to prefer to sort the passes in different folders, after the render

http://www.linuxjournal.com/article/8919
got me going with a rather simple script.

So I came up with this ultrasimple script:

find . -type f -print |
while read -r pathname; do
   
    filename=${pathname##*/}
        
    # very unclean because hard-coded 0000.exr ending but good for now
    new_dirname=${filename%????.exr}

    if [ ! -d "$new_dirname" ]; then
        mkdir -p "$new_dirname"
    fi
    mv "$pathname" "./$new_dirname/$filename"

done

Samstag, 15. Oktober 2011

OpenEXR and R13

I am changing my pipeline to accomodate OpenEXR for the following advantages:

  • All channels in one file (no messy directories)
  • 32-bit 
  • lossy compressions available 
These are the downsides
  • You cannot re-render just one channel (ie a mask), but from experience I have rarely only rendered one channel, plus it looks like C4D would have to render to whole image anyway, so not a lot is gained. For special cases, such as masks I will generate subdirectories and they are highly compressed anyway.
  • No 16-bit support in R13 (maybe, they'll bring it back) - so the only option is to go with 24bit lossy, which according to forums will just truncate 32->24 bits and rezip the file. For my purposes this is more than enough, as I don't layer a lot of stuff.
  • ExtracToR in After Effects is rather slow, but this seems to be related to the file transfer.
For my test frame, I was faced with those sizes (I didn't measure multichannel here)
  • All compressions RLE, ZIP, ZIP16, lossy PIZ wavelet, etc around 20-24 MB
  • The 24B format comes in around 12 MB (ie half)
  • A 16-bit PSD file comes in at around 12.4 MB - so 24BLossy is the way to go.
  • 8-bit DPX is 6.2MB
So with 24 bit lossy format I gain more than enough control. 

Freitag, 14. Oktober 2011

Copying mograph simulation to high-poly object

It is not straightforward to bake a simulation in an array of dinamically simulated mograph objects. This is my way of doing it:

  1. Bake the simulation in the cloner (it doesn't matter if the dynamic tag is on the cloner or the object)
  2. On the final object create an expresso tag and add a node for the Cloner object and the Final Object.
  3. Add mograph data node and connect global matrix to the Final Object global matrix.
  4. Step through the index manually until you've arrived at the object.
Alternatively, it might be easier to find the object index with Mograph selection, but I couldn't figure out where the information on the selected mograph clone index would be displayed.

Mittwoch, 12. Oktober 2011

Lighting Setup and Balls on Table

 First lighting setup - the balls should be able to pop from the background, so I've added a white reflective band from the outside which helped a lot. Might be slightly too dark, but as looks rather realistic.


Pool Table Modeling

Added another 2 hours. Still working on the lighting of the whole thing. Torus is added for shadow measuring and color bounce.