Originally posted by mulenmar
View Post
VP8 and VP9 in libvpx have a real-time mode, where it is as efficient as possible when encoding the stream in real-time, however, compared to Theora both have the potential to be more efficient at the same speed Theora encodes (if somebody puts the effort in to run the encoder down to Theora level). The main problem with VP8 and VP9 is that we're more or less stuck with libvpx, which sadly doesn't squeeze as much from those two format out as it could (also for the real-time use case). But then we have some VP8 and VP9 hardware codecs available, which can be used for real-time now.
Looking ahead at AV1 I think we'll see some really great real-time encoders (rav1e for example). AV1 also has some coding tools that were added explicitly for screen content, so the efficiency could be drastically better than VP8 and VP9 (also bigger transform sizes help a lot with screen content, with many big areas where the same color and CDEF - deringing filter should help sharp edged areas - like text).
Comment