Comment by k03im-d

if this hadn’t been performed on a real, live, over the air signal with a real, live, audience providing feedback to the station, we (and my competition, other broadcast brethren) wouldn’t be talking about this now.

fyi - AVC and HEVC decoders are, sadly, far more fickle than they should be, in 2024. we would like to believe they all implement a working HRD buffer model, read and obey stuff we put into the SEI and VUI NAL units, as well as properly support basic functions like slices and layers (such as non-ref P frames for 60P refinement of a 30p sequence), and etc.

heck, based on current feedback and testing, the least common denominator for maximum reference frames in AVC is four, and HEVC is five. sad, but better than mpeg2.