Re: Alphas, mattes, adds and overs

["Axel Akesson" <axel.akesson(at)gmail.com>]

I won't get into this discussion. But I like Shake.... ;-)

Add it up and then grab the alpha from the beauty pass, it's painful to do in AFX though, much easier with nodebased (I told Al how to do it in Shake).

On 6/26/07, adrian <adrian.wyer(at)fluid-pictures.com> wrote:

oh please guys, i've been using afterfx for years, and i can still do stuff that flame guys don't get.... it's a great tool!

but, i digress......

 

we tried compositing with separate mattes (both rendered from 3d as passes and pulled channels from the RGB renders)

compositing using track mattes creates the same issue, as it does when using composite with matte in FX tree

 

what i find weird is, we have never seen this before, i was tempted to lay the blame with mray, but maybe it was there all along, and

only with us having to work at full HD all the time, does it now become apparent

 

hmmmmm

Adrian Wyer
Fluid Pictures
33 Glasshouse Street
London
W1B 5DG
T: +44 (0) 207 434 5133
www.fluid-pictures.com

----- Original Message -----

From: Axel Akesson

To: XSI(at)Softimage.COM

Sent: Tuesday, June 26, 2007 5:19 PM

Subject: Re: Alphas, mattes, adds and overs

When will people understand that AE is for motion graphics and not vfx comp...

On 6/26/07, Robert Chapman < tekano.bob(at)googlemail.com> wrote:

yeah, bin the AE guys, its not *that* complicated to think about now is it, they should go with what works and stop whining.  :)

on a more constructive note, perhaps its possible to isolate out the 'add' part in the alpha channel only and use a track matte - which might not confuse the poor dears so much. bless.

 

On 26/06/07, Alan Jones < skyphyr(at)gmail.com> wrote:

Hi Guys,

Got a production question for you as I'm a bit stuck coming up with a
"good" solution (read solution artists find acceptable) to this issue.

Say you run a couple of objects in a couple of passes/framebuffers and
the block each other with 0,0,0,0 as you do. This means along the edge
you'll have pixels with an alpha of n for object a and 1-n for object
b. Which would be "the right answer" (tm).

You then go to composite this. Artist pulls an over node and of course
this winds up with slightly transparent line running along the join.
If you use an add (and it adds the alphas as well) then you get the
result the artist expects (between the two objects we know it's solid,
but viewing either individually that's not the case), but of course if
add was used in a scenario where they was overlap then you'd get glowy
grossness etc.

The guys don't like the idea of using an add because it seems counter
intuitive to them. Where they'd just use an over if it went over a
plate (of course - the plate is solid so you really are going over).
It's not a huge issue (if you ask me anyway ;-) - for node based
compositors) because you just use the add node where it's required and
over where it's needed, but it's freaking the After Effects guys out
because it apparently means subcomps and other things to obfusificate
the comp when they're looking at it.

So - was that obscure and long-winded? Anyone got any comments/suggestions etc.

Cheers,

Alan.
---
Unsubscribe? Mail Majordomo(at)Softimage.COM with the following text in body:
unsubscribe xsi