[Steve Bentley]

Can you fill us in on why you need the mask? If its to do the color change, you might be better not trying to isolate the sweater perfectly with a hard edges matte: that’s going to be tough.
I think even if you knitted the sweater out of green screen material it would still be tough pull. There are shadows, the hair is not really that different tonally from the sweater and background – its really a one tone poem going on there.
There is not going to be one technique that works here. Lots of roto, lots of garbage masks, lots of layering back in things like hair with transfer functions, lots of tweaking multiple soft edged keys.
I think a reshoot (or since this is stock, I think, just a “shoot”) with a contrasting color for the sweater would give you more to work with.
I’m always amazed at Mocha, its pure Voodoo sometimes, but it too has its limits as Cassius rightly points out. I can’t believe it would be any worse than doing it by hand as Mocha should take some of the repetitive shape differentiation off your hands, if it can’t do it all. Remember, you don’t have to do the whole sweater as one piece – you can stitch all the bits together on top of each other in the end.

[Steve Bentley]

Also that motion blur looks pretty steppy – go look at one frame. Its pretty hard to motion blur after the fact because you would have to blur each object separately because they are moving radially out from the camera center and rotationally and laterally as the camera weaves around. A radial blur set to zoom would get you closer.

But as Mark pointed out, Twixtor or RealSmartMotion blur can figure out every object’s own blur because it compares frames pixel by pixel. And this scene is the perfect use: Twixter and RSMB have trouble with objects suddenly coming into frame since there is no “before” frame for that object to compare things to. But this shot has everything in frame to start and it all leaves frame eventually.

If you did this for real with a 180 degree shutter, things would be really really really blurry at the edges (think: unidentifiable smeared objects as a still) and less so at the center. So a global blur doesn’t always work.

When we do this for “The Flash” we also add little trailers at pointy edges and streak those points a little longer You can either make a mask or just let a crushed luma key pull out some details that you streak longer than the radial blur for the whole scene. (if that’s what you’re going for). It really adds a sense of speed and covers other sins.

[Steve Bentley]

The problem with traveling masks is its really hard to bend and turn them without affecting stuff that’s close along the way, especially when you change directions. Its also a lot of keyframes.
Instead, try either of these.
On a separate layer above your scene:
You can make a painted path that has a gradient from one end to the other (either in AE with the new gradients on shapes, or in photoshop with a brush that changes color over its stroke – so black at one end and white at the other gradating in between).
The use the compound blur effect and feed the gradient as the “blur” source and it will act like a mask that runs along the length of the path proportional to the Maximum Blur in the effect. Use this element as a mask to reveal a glowy element. Then use a second gradient with compound blur to follow behind and close it up (assuming you want that)

Or

Use the stroke effect to follow stroke the path and animate the start and end parameters of the effect to give you a moving hot spot in the middle of the stroke – blur and glow as needed. You can either do this on transparent so the background shows through or make it a second layer and use a transfer blend like “screen” or you can do both of these compositing techniques.

[Steve Bentley]

And don’t forget the screens themselves. Color gamut settings, black level settings and backlight color temp (both LEDs and LCDs) can all have an affect on what you have to deliver. The onsite guys love to point fingers back at the content creator when in fact it’s usually them not knowing their own poorly setup hardware. You really have to get in there and investigate the playback and screens aspects to make sure WYSIWYG.

[Steve Bentley]

That’s a pretty simple gradient. Why not start from scratch and make two gradients you can apply a ripple or wave to. Then you aren’t fighting the built in ripple. From there you can animate the wave, the colors, the position or offset of the colours and it’s a simple comp over you man footage.

[Steve Bentley]

AE is a compositor. Why not compress outside of AE? Quicktime pro does a fine job(though not the best), but for my money ffmpg is crazy fast, completely flexible and has no limits except the overall 264/265 specs. We’re routinely doing massive sizes that would make AE cramp. Did I mention ffmpg is free?

[Steve Bentley]

I hear you. I’ve been complaining for years about the ever increasing screen resoluions and the ever decreasing size of icons and my againg eyes.
In this case I find thinner the better (for accuracy). I even turn off the “never aligned” bounding box.
But i’m pretty sure you can’t thicken it up. What about changing it to a bright yellow green- it’s the color the eye sees best. (think safety vest color)

[Steve Bentley]

Hey Eric, could DaVinci really turn that sweater, say, deep red?

Stefan, as for Roto – try mocha. It would at least give you a hand with its smart shape detection. Normally when we know we are going to do the color shift thing we shoot the talent with some unique color sweater so that we can simply either warp the hue or replace the color. Any time you can get away from keying in this type of application it produces better results. (rotate hue instead of keying sweater or bg).

And here’s an old Color Corrector’s tip. If you are going to roto, do the roto on a separate copy that sits on top of the original, warp the color of that one and let it sit on top of the old one. This clip is perfect because the sweater is gray. When there are small “leaks” in the roto mask they won’t be as visible. Back in the day when touching up color photos we used grayscale paints , not color- in small areas the eye is better at tone than color so it just assumes the small gray splotches are the same color as the surrounding things.

[Steve Bentley]

Referencing Lars’ comment, it, because its a video tutorial, perhaps the resolution of the screen was reduced for capture and the compression is making the lines a little thicker. You can also make the background color darker which would help with contrast. Is there a reason you want them thicker?

[Steve Bentley]

There’s some clues in your post that makes me think you might not be approaching this correctly. The font size is determined by the text object parameters so copying the text from another document shouldn’t be affecting the outcome in AE. But there might be other formatting information that is coming in with the copied text (leading for instance) that might push the text “out of view” in a single line text object. Can you upload the project? (just drag it into the reply box here)