[Keith Mcgregor]

@darby: Yeah, it’s been my experience when keying for web based videos and mixing with progressive bkgrnds I get a better look. (comes from not exporting ntsc from AE) But this was when I was doing DV25 SD from a panasonic DVX100 and going out to the web so the end result isn’t interlaced. And just recently I was helping someone out with sony xdcam footage which was interlaced and when they keyed they were getting black fringes on their mattes and when we turned separate fields off it was gone, if we used matte choker it would sometimes increase the density of the fringe. But I dunno, try it yourself and if it comes out fine then don’t listen to me, keep with your workflow. Oddly enough I can’t find any footage to test with, I seem to only have progressive footage lying around, yay for the future!

@christina: I forgot to mention when I key I never use the final result, I use intermediate, and I always leave it unpremultiplied. I take care of the spill, edges and color outside the keyer and add a little wrap if needed.
-Keith

Reality? What did you make it?

[Keith Mcgregor]

(if this is a repeat post its the cows fault for kickin me off after a long-winded explanation)

Ouch! I would definitely get her isolated, key a hard core matte, place that on top of the original footage, change the color of the core matte, bring them both into a comp and key that comp to use as a track matte. Hope this helps.
-Keith

Reality? What did you make it?

[Keith Mcgregor]

Wow! Ouch! In this case I would make her a separate layer just to treat it and key it. Try using color balance to remove some of the green highlights or use the core matte on top to ‘hold out’ the color effects for the rest of the footage, to reduce the chroma on her while leaving the bkgrnd green. I didn’t get such a bad result for 10 min worth of work, but I don’t know what it lookies in motion. You are keying in 16 bit right? You should always be at 16 bit when keying and rendering out, as it does a better job of reducing banding and retaining color information.

Now that I look at it, is your footage interlaced? It would be best if you tell after effects there are no fields to interpret (separate fields off in interpret footage dialogue tab) while you key, as the resulting fields make edge chatter, pop and swim. If you can, tell the cam operator to use a progressive setting whenever shooting green screen, and to light better. (I know. I know, camera operators know everything…but you could still try?)
Hope this helps.
-Keith

Reality? What did you make it?

[Keith Mcgregor]

Are you using color management? If so what is the setting? What colorspace is the clip being interpreted at? What bit depth did you export out to?
Did you try a test render of a second or 2 where the spots show, and do they end up in the final render?
-Keith

Reality? What did you make it?

[Keith Mcgregor]

In my opinion there is never a need to use HQ as it just makes the file sizes too large wasting disk space, and there is no benefit you could gain when transcoding 5d footage to it. I would recommend following canon’s suggestion that you use LT. If you do a side-by-side comparison of HQ and LT, do you see any perceptive degradation of the footage? I have not yet been able to, and as you say the data rate is more than enough, as we see here:
for the canon 5d-
HD bitrate is approximately 38 Megabits per second (4.8 Mbyte/s); and: The 21 megapixel sensor is
downsampled to HD resolution by only using every third line and 4:2:0 chroma subsampling
from: https://en.wikipedia.org/wiki/Canon_EOS_5D_Mark_II

And since we also know the chroma sampling is a mere 4:2:0 what could you possibly gain? Now if I had a 2k image with full color sampling I might think of pro res 4:4:4 but then again the original footage needs to be captured at that kind of sampling from a camera that can actually do it.

As far as being in the same timeline, you should be able to mix all flavors of proRes 422 and not have to render except for final output.
Hope this helps.
-Keith

ps- if someone can actually show me a visible difference you are welcome to try, or if you have scientific data to back your claim please let me hear it.

Reality? What did you make it?

[Keith Mcgregor]

Well, first pre-treat the footage for better contrast and better greens, trying to remove some green from your foreground footage. If you need to, make masks for certain areas and use adjustment layers for those isolated areas. Then look at how noisy your blue channel is and use the ‘remove grain’ effect only on the blue channel to soften the noise, try not to blur the hair too much, name this comp “preTreat”. Take that comp and place it in a new one and pull a hard, core matte. Label that comp “coreMatte”. Drop “preTreat” on the make new composition button and pull a softer, edge key in there, call this comp “edgeMatte”. If you need to make more garbage masks do it in “preTreat” so it is in all of your comps (removing c-stands, tracking markers, uneven screen parts, and on…) In the edgeMatte comp, refine it for most of the edges and if you can get a good edge around the hair, then good. If not, dupe the edge matte layer, roto a mask around the head or just the problem area, and pull a softer key for that- just concentrating on the hair. Feather the mask a little and you should be good. Use option=4 to see your alpha to make sure it looks good, option=4 to return to the RGB channels. Now after all this, place the “edgeMatte”, “coreMatte” into a comp and name this comp “alphaMatte”. Take the “alphaMatte” comp and your original footage into a comp name it “key” and use the alphaMatte comp as an alpha track matte for the original footage. Now you are free to apply spill suppression, color correct, any transformations and match with the background! Vóila!
Hope this helps. Cheers!
-Keith
ps- please post screen shots in the future when asking for help so we know how to help you better

Reality? What did you make it?

[Keith Mcgregor]

Look for some reference in stock photos so you know where and how the highlights fall. Then create a mask for the eye(s) that need the water effect, create a solid and apply wave world effect to it. Find or make a ground for wave world, (I usually use a clouds filter in photoshop) and get a high contrast image with levels effect and adjust accordingly. The reason for this is just so we can get some distortion for the water on the eye. Get a good sim with high samples and low wave speed and find a good, small spot where the waves are moving up or down (you can flip it if there is a good reaction but it’s flowing the opposite direction). Then use the matte for the eye(s) as a luma or alpha matte for the wave world solid and adjust the simulation until it’s to your liking. If you also place a displacement map with light settings on a duplicate of the eye mask and have that layer as an add blend mode to pop out some highlights. Some finessing will be needed on the wave world solid to get it moving correctly but this should work. And of course if the person is moving you will either need to track or manually place the crying effect, but have it precomped at this point so it’s just a single layer, much easier to move without ruining or parenting.
Hope this helps.
-Keith

Reality? What did you make it?

[Keith Mcgregor]

Are you sure you need the cam to move and can’t have the earth move? 99% of people watching it won’t know the difference. But yes, just place your current time indicator at the first keyframe and select the property ‘position’ (the actual word, not keyframes) and copy. Select the camera’s position property and paste, then select earth again and click on the stopwatch to remove the key frames from earth layer and it should be good to go. I did a test with a solid and a cam and it worked fine, kept the same timing and movement, but if it doesn’t look right then just redo the cam movement, practice makes better!
-Keith

Reality? What did you make it?

[Keith Mcgregor]

The easiest way is for them just to replace the footage by right-clicking on the file in the project tab and choosing “replace footage” and it places it where the previous footage starting point was, if it’s a shorter duration then someone has to manually reorganize the timeline, but it will be placed at the start of the video that was replaced. It’s not hard or time-consuming to go through the timeline and use the i, o to get to the layer beginning and end points or using left/right brackets to place the layer at the playhead position, very simple and should take no time at all (depending on how many videos I suppose, and how organized the timeline is).
-Keith

Reality? What did you make it?

[Keith Mcgregor]

Make sure you are looking at it top down and when you hit the switch, see where in 3d space it lives. It could be behind your camera (are you using a camera?), and cycle through the custom views and orthographic views to see where it’s placed in 3d space. If your original comp has the continuously rasterize switch on and you can scale it p do you really need the precomp? Make sure your original is not 3d enabled, the position for Z is 0 and the position for your precomp is also 0 for Z axis. I am not seeing the same thing you are, when I hit the collapse switch my object stays where it is and doesn’t disappear. When you hit the switch it takes into account of where it is in 3d space in the original comp, so maybe you moved it?
Hope it helped.
-Keith

Reality? What did you make it?