- This topic has 4 replies, 5 voices, and was last updated 12 years, 5 months ago by .
-
Posts
-
Hi Guys,
I was just wondering, would it make any difference to my keys if I do it before or after plate stabilisation? I have fine blonde hair detail (in this particular case) against green (I know, less than ideal) on 2k res, and was wondering if some of the finer adjustments that the stabilization process goes through (sub-pixel positioning, etc) would adversely affect fine details?
On the other hand, cleanup with roto is much easier on a stabilised plate. Or do a key in one level of the comp, then stabilize in another level, then finally roto over the stabilized/keyed footage to clean up? Or is that overkill?
Thanks so much,
Scott.u didnt say which comp software you are using for keying / stabs ?
On a pixel level, stabilizing would create sub-sampling, which will mix the edge pixels by half a pixel. So u should be better off keying first and then stabilizing to achieve pixel accuracy.
If you could prevent the stabilizer from using values with decimel points, you can prevent sub-sampling.
You could key first, generate the mattes then stabilize and apply the stabilized data to the matte layer. This would negate the problem of keying on a already sub-sampled image.
ps-make shure you are keying on a RGB pass to achieve this level of accuracy. if you are trying this on graded footage, u have already lost pixel accuracy
that’s one of the potential gotchas of comp software being so manual – in the wrong hands you can compromise your image quality pretty quickly. in this case it’s a good thing you asked!
a good acronym to remember in cases like these, to help keep comps in order, is – METL. (*cue judas priest shriek and glass breaking here)
your comp should follow this route.
first – M
Masks. mask off your footage first (where possible) for work downstream. apply your rotos early and make use of that DOD on footage you’re lopping off.E
Effects. apply your effects next. your keying, glowing, lightsabering, etc.T
Transform. this is where you move and scale your layers, because doing so before this point risks subsampling or pixel smoothing prior to effects being applied… which would be baaaaad.L
Layer. do the deed, comp the comp!and of course there are alwaysexceptions, and you can get away with things like color correcting and ending ops like regraining further down the chain. and I s’pose there are times when you need to apply effects to precomps of several layers. in which case, just METL on those as if they were virgin layers.
ya know… generally following this path is a good habit. just keep METL in the back of your brain, and everything will be as righteous as a Randy Rhodes guitar solo…
I agree the rego’s answere… its only the best way…
- You must be logged in to reply to this topic.