Coding efficiency and rate control

A CBR-against-CBR comparison mixes two things: how well each encoder codes, and how much its real-time rate control costs it. This run measures them apart, on the development clips: each encoder's CBR curve and a fixed-QP curve over the same rates, for our Medium (default) presets and for x264 medium at both tunings.

Development clips, one build. These are the clips we tune beside, not the held-out set, and the run has no low rungs. A clip counts only where the curves overlap reliably: the number of such clips is in brackets. In this low-delay mode (one-second buffer, no lookahead) x264's own rate control is expensive for it, which is part of why our CBR margins are larger than our fixed-QP ones.

What this shows