[Andy Zou]

My off-the-top of my head guess is that it takes a second to decode the h264 and put it into ram or some sort of cache or preview, which is why a cold start on a clip may take a second but giving it a moment to load it may not.

[Andy Zou]

Like I said, guys, I know what VBR is. Perhaps my question is beyond the scope of a forum and I’d need to read up on h264 encoding.

I wanted to know how it decides where to allocate its allotted bits for compression:

Judging by how low-bitrate compression often looks like blocks, I’m guessing the algorithm partitions the frame into rectangles, and smaller rectangles within those rectangles until it isolates partitions that have movement. But does an algorithm decide between preserving finer detail or does it distribute its compression “equally”?

How does a VBR encoder decide that it “needs” to tap into its peak bitrate, etc?
I know with 2-pass encoding it’s:
1-pass analyzes the clip
2nd pass encodes the clip.

How does the 1st pass analyze the clip? What’s the methodology there?

[Andy Zou]

I’ve found an inelegant solution, but I don’t like it and am welcome to other options:
I created two more duplicates of the Duplicate Potato, and created copies of our shadows, and alpha matted them thru an inverse of the duplicate potato. But this leaves some fringing on the edges that may be noticeable to no one but myself :-/