PDA

View Full Version : p_motion_to_RGB in combo with Nuke 4.7. Continued discussion from the "other" thread.


NEO^AMiGA
09-27-2007, 09:23 AM
I have set the shader up as in the doc linked to below. But still doesn't get any result. Have I missed something obvious?

Sorry for external link but here is my setup:
http://www.irry.com/neo/bilder/temp/3d/nothing.gif

As I understand mip_motion_vector don't support nuke motion vectors format, Master Zap please correct me if I'm wrong.

But if you need it, you may use some another shader, for example my p_motion_to_rgb, it work fine with Nuke (and possible with new vector blur in DF 5.2).
You may find tutorial in docs in archive or online version here:
http://www.puppet.cgtalk.ru/shaders_p/docs/
direct link to 2d motion blur tutorial here:
http://www.puppet.cgtalk.ru/shaders_p/docs/2d_blur_motiontorgb.html
shaders pack you may dowload here:
http://www.puppet.cgtalk.ru/download/shaders_p_e.shtml and here is updated dll for Maya 2008 (win32 and win64 only):
http://www.puppet.cgtalk.ru/download/shaders_p_3.1.1_maya2008.zip

Sorry for offtopic.

Puppet|
09-27-2007, 10:14 AM
Are you sure that your vectors is floating point? Value may be negative or more than 1.
Format in p_motion_to_rgb should be 'Raster Space'.

For getting full floating range format I recommend you use Maya Batch.

Don't forget enable 'Alpha' in Nuke VectorBlur node.

NEO^AMiGA
09-27-2007, 10:45 AM
Oooh! You are so right on spot. My bad. Didn't render float. And i missed a couple of lines in your documentation. Didn't see the line about "Raster Space" at first. Now it seems to work and it seems to work damn good to!

http://www.irry.com/neo/bilder/temp/3d/everything.gif

Just tested on proxy beauty layer now so it looks a bit rough. But this is probably the shader I'll go with. I could prolly go with the mip_motion_vector that we have discussed in the other thread. But now I have this working and I would prefer to render from 8.5 in this case.

You should hang in the Nuke forum over at vfxtalk.com dude. =) Thanks a lot for helping out! You saved my day.

Are you sure that your vectors is floating point? Value may be negative or more than 1.
Format in p_motion_to_rgb should be 'Raster Space'.

For getting full floating range format I recommend you use Maya Batch.

Don't forget enable 'Alpha' in Nuke VectorBlur node.

NEO^AMiGA
09-28-2007, 03:44 PM
heya!

Everything is not sunny over here. =/ Have some issues. Look at this picture (http://www.irry.com/neo/bilder/temp/3d/artefacts.gif) (pretty large pixelwise...). When I enable the alpha I get artefacts on some frames. I have measured the colorvalues in the vectorpass and from what I can see it's ok. Rendered as 16bit half to exr. Will try with 32 bit and see if there's any differece... But as I said, can't see any oddness on the seperate passes when i measure in them.

I think the picture is pretty selfexplaining. Tell me if it's not.

Anyone with any suggestions here?

One thing to have in mind is that the coins here are what i call "proxys"... Not rendered with the "hires" mentalray material but instead just rendered with a simpler shader and maya sw.

Are you sure that your vectors is floating point? Value may be negative or more than 1.
Format in p_motion_to_rgb should be 'Raster Space'.

For getting full floating range format I recommend you use Maya Batch.

Don't forget enable 'Alpha' in Nuke VectorBlur node.

NEO^AMiGA
09-28-2007, 03:54 PM
I added another picture (http://www.irry.com/neo/bilder/temp/3d/artefacts2.gif) that illustrates the artifact.

NEO^AMiGA
10-09-2007, 08:31 AM
After rerendering the vector passin 32bit float instead of 16 halffloat things seems to work. Though I tested this shader for another render now and I when I look in nuke the alphafrom the colorpass does not align with the vector pass. Or rather... The vector pass does not align with the colorpass. It's slightly offseted forward in time it seems like. Some kind of shutter option that's not correct?

Puppet|
10-09-2007, 08:34 AM
It's work fine with 16bit half float(OpenEXR) and with 32 bit float(TIF or IFF).

Are you sure that all parameters are same as in tutorial said?

NEO^AMiGA
10-09-2007, 08:52 AM
Hey Puppet.

These are set:
- Shutter Open = 1
- Shutter Close = 1
- Custom Motion Offset = On
- Motion Back Offset = 1

But in addition to that all the settings that has to do with motionblur isn't in the tutorial. Sooo.. My other motionblurrelated settings are:
- Motion Blur By = 1
- Motion Steps = 1

-Static Object Offset = 0
-Time Samples = 1

-Time Contrast = 0.2, 0.2, 0.2


It's work fine with 16bit half float(OpenEXR) and with 32 bit float(TIF or IFF).

Are you sure that all parameters are same as in tutorial said?

NEO^AMiGA
10-14-2007, 09:56 PM
Hey. Sorry for nagging about this. Just wondered if Puppet| had any additional info about this? =)

Puppet|
10-15-2007, 08:35 AM
Please share simple scene.

NEO^AMiGA
10-15-2007, 11:40 AM
hey man

You can find a rar'ed scene here:
http://www.irry.com/neo/bilder/temp/3d/p_motionblur_test.rar

Render the vector layer and color layer at a frame around 300-400 and compare them.

Thanks for helping out man.

Please share simple scene.

Puppet|
10-15-2007, 11:58 AM
You have 'Motion Back Offset' = 0.5 instead of 1.0 in your scene.
It cause of motion vectors offset.

NEO^AMiGA
10-15-2007, 12:01 PM
I notices that to after that I had posted the first note here. But I am pretty certain that I corrected that and rerenderd it all... But there was a lot going on that day so I might have missed it. I will rerender with the correct setting and see. =)

Thanks puppet!

You have 'Motion Back Offset' = 0.5 instead of 1.0 in your scene.
It cause of motion vectors offset.

JulianS
11-20-2007, 04:39 AM
I dont know why i am having a hard time with this one, it should be fairy easy to setup.

can someone please give me quick hand here is my setup
thank you

http://creating3d.com/Help/MotionVector.jpg

JulianS
11-20-2007, 04:56 AM
So here is my solution I remove the Alpha info because i am only rendering RGBs no RGBA so that look like it did the trick,

it still looks a bit funky I wonder if that if that is the result ???? :shrug:

http://creating3d.com/Help/MotionVectorFix.jpg

Puppet|
11-20-2007, 08:05 AM
Sorry, but what is the problem to render correct alpha? Correct Alpha needed for motion blur.
Or I have missed something?

JulianS
11-20-2007, 06:23 PM
Acctually i have no idea but went i did the renders i use RGB (float) 3x32 bit as my data type in the frame buffer for the vector and diffuse pass.

So i should assuming my diffuse pass should be RGBA so it has the alpha, correct!?

CGTalk Moderation
11-20-2007, 06:23 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.