Anybody using reelsmart motionbur with vectors generated by cinema

[AdamT]

It's a bit tricky. You have to render out the vector pass *without AA,* then render the scene and object alpha(s) *with* AA. In AE, you'll precompose the vector pass with the object alpha and use that as your RSMB vector pass.

[chris_b]

It is a little bit tricky because of a mismatch issue with C4D's motion vector output and the RGB or Buffer alpha output. I have posted requests for this bugfix on the Maxon site a few times now. There is a good document on workflow here:

http://www.revisionfx.com/support/faqs/hostfaqs/cinema4d/

[chris_b]

Yes.... Adam is absolutely correct, you will also have to render twice since even though the Motion Vector pass appears not to be antialiased, outputting it at the same time as the RGB pass causes problems with the data (incorrect output of the Motion Vector pass).... This is also a frustrating bug as you would in most cases want to output the passes at the same time as the RGB.

[v-empire]

Hmm, sounds like wasting the time for not calculating the moblur inside cinema with rendering in two passes. So no timesaver any more.

The case that you mentioned sounds to me like only bluring certain objects (combining object alphas and vector data).
What if I want something equivalent to scene motionblur.
That means no alphas just vector data for the whole image?

Thanks for your help
Holger

[v-empire]

Wasn´t there a little plugin somewhere before the vector buffer was introduced to cinema?
Would this do the trick in one pass.
I think vixol was the name (c4d 10 compatible?)

Thanks
Holger

[rsquires]

vixol was the bomb and provided a vector pass correctly in that it didn't antialias the edge but provided an unmultiplied image. Doesn't work any longer unfotunately with R10. I think the last version I had it on was 8!

It's something that Darf cooked up and knowing what problems were had with Jenna being redistributed etc I don't imagine it ever coming back any time soon.

So if anyones listening and knows how to do this properly it would be a great addition because the cinema implimentation is flawed.

However I find Reelsmart to be so good that it pretty much works without a vector pass. It's not %100 accurate and does have a few issues but I find these are outweighed by it's speed in acheiving a result as opposed to rendering mo blur within C4d.

regards

rich

[Blabberlicious]

It is a little bit tricky because of a mismatch issue with C4D's motion vector output and the RGB or Buffer alpha output. I have posted requests for this bugfix on the Maxon site a few times now. There is a good document on workflow here:

http://www.revisionfx.com/support/faqs/hostfaqs/cinema4d/


BUMP

Is this still an issue?

[chris_b]

Yes this is still an issue....

You can use a matte choker to help clean things up... Also.... the vector scale in
the render options is key.... the default value of 128 is often too large.

[Blabberlicious]

Yes this is still an issue....

the vector scale in the render options is key.... the default value of 128 is often too large.

Thanks for confirming that.

Considering the multi-pass goodness that maxon promotes (rightly), I find it a tad perplexing.

So C4D lacks a true coverage buffer, like Lightwave?

Does that explain the lack of surface ID export for RPF/RLA?

[Simon Wicker]

i've always wanted to use reelsmart to test motion blur but the added complexity always put me off.

i've had a little play around and i'm trying to understand the root cause of the problems people are having.

is this to do with the mismatch between the velocity multipass and the alpha channel cinema saves? the velocity pass is rendered without AA while the RGB is rendered with AA so the edges no longer line up correctly?

would the solution be to render the velocity pass with AA and then un-premultiply it (so that the edge is expanded out over the edge of the alpha channel)?

i know that vixol was said to have worked better at this and i wonder what they were doing right and cinema doing wrong.

cheers, simon w.

[Hilt]

I've settled on doing everything MB related in compositing without hassling with the C4D motionblur.
MB in C4D wasnt good in R8 and it definitely hasnt improved since then. I totally ignore the MB in C4D and just do it all in the post without bothering myself with vector MB or AA or stuff like that.
Render the layers and do the MB all in compositing. It's not 100% accurate but it's 99% faster.

Darf could do the good deed of the day and release the code so Vixol could be translated to R10..

[rsquires]

Depth passes and vector motion passes by their very nature have to be non anti aliased.

C4D's implimentation is right.

Take the depth pass for instance. Depth is calculated by levels of grey, so if one anti-aliases an edge for instance then you run into problems because it is putting in levels of grey that should not be there. This is the same with a vector pass, so it corrupts the motion blur.

But.... and here's the kicker, in practice this is a pain. I have just got zblur 2 and it allows for an anti aliased depth pass. And guess what. In After Effects with the Frischluft plugin it looks fantastic. Admittedly the project isn't a difficult depth of field calculation, ( ie there aren't things passing close to camera) but it still looks terrific.

Hi Simon,

when I used to use Vixol it was superior to the C4D non antialiased output. It rendered the motion pass with edge anti aliasing and gave a straight alpha edge into the bargain. It also rendered anything within the alpha with antialiasing as well, which is something C4D doesn't do. Please Darf I know you're not listening anymore but this was a life saver.

In my experience with Reelsmart Pro 3 I often forego the motion pass because it is actually totally useable without it. Just download the demo and see for yourself. Yes it isn't physically correct and in some places it can act weird but for Motion Graphics work it is often a lifesaver.

Here's something I tried the other day. I was testing the P2 camera that a friend of mine has and asked him to shoot some 1920x1080 bluescreen with it. I was just using Apple Pro res but got him to shoot at 1/250th shutter speed which cut out pretty much most of the motion blur. I just shook my hand and a furry mike cover in front of the camera. I was able to get a cracking key because I didn't have to deal with trying to key blur, and here's the best part. I was able to put the blur back on with Reelsmart . When I next shoot blue/green screen I am going to try and use this workflow as it yielded some very impressive results.

I do wish Maxon would allow at least the option to antialias the vector pass or maybe write a shader that we can use instead.

all the best

rich

[Hilt]

1/250th shutter speed

OT, but it always involve heavy debating with DOP whether to have motion blur in edges or use higher shutter speed and simulate it in post production. I usually have my way because post production costs more than using proper shutter speed..

[rsquires]

OT, but it always involve heavy debating with DOP whether to have motion blur in edges or use higher shutter speed and simulate it in post production. I usually have my way because post production costs more than using proper shutter speed..

I'm the one who has to deal with it so I am sure I'll be able to sway them.

rich

[Simon Wicker]

thanks for the information everyone.

i'll have a play around and see what results i can get from this.

cheers, simon w.

[chris_b]

Hey Guys...

My experience with Vixol back in the day was also far superior to the inbuilt Vector Pass. Vixol had a toggle that allowed you to unmultiply at rendertime so that you always had the coverage you needed in the Vector pass. I would then use the RGBA alpha as a track matte for the unmultiplied vector pass. The nicest thing about this was that it could be output as a multi-pass layer at the same time as the other passes without the necessity to set up a separate pass with AA turned off.

[CGTalk Moderation]

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.