[infofarmer]

you'd have to "rebase" all the other frames to be derived from those

[coryrc]

You wouldn't have to any more than you need 0 through N frames in memory to calculate frame N+1. Whatever your decoding state completes at frame N can be considered a key frame.

[infofarmer]

decoding state can be bulkier than a key frame and opaque to the CPU if hardware acceleration is used

I wonder if caching semi-compressed frames would be more efficient in either case (CPU or GPU)