Codec thread: Corrupted segment_ids edition

Here are some more comparison images along with encoding speeds.

cpu-used=0 | 1.33 fpm
cpu-used=1 | 5 fpm
cpu-used=2 | 12 fpm
cpu-used=3 | 18 fpm
cpu-used=4 | 25 fpm
cpu-used=5 | 37 fpm
cpu-used=6 | 33 fpm
cpu-used=7 | 35 fpm
cpu-used=8 | 39 fpm
Output filesizes range from 3103562 bytes at cpu-used-3 to 3223015 bytes at cpu-used 8.
How fast does it encode on a TALOS II?

Attached: AV1_cpu-used-6.png (640x480 285.33 KB, 282.05K)

Attached: AV1_cpu-used-8.png (640x480 292.78 KB, 285.55K)

>>>/bog/

Could a full VP9 decoder or encoder (like in the Kaby Lake iGPUs) work as a hybrid AV1 decoder/encoder with a software update (like the VP9 decoder/encoder in the Skylake iGPUs)?

cpu-used 0 in VP9 is for extreme qualityfags only. 1 is a lot faster with only slightly lower quality, while everything above 1 is inefficient (should just go ahead and use x264 instead if you were going to use a cpu-used value above 1)

A partial HW decoder via software update might be possible as AV1 is sort of an extended fork of VP9, but any sort of quarter-decent AV1 HW encoding will definitely require a custom ASIC.
Judging by how unreliable the current reference decoder/encoder are we're probably at least 2-3 years from these sorts of thingsgiving MPEG and friends more time to sort out their patented HEVC licensing faggotry.**

Try opening a h.265 video with a browser (like most non-Zig Forumsfags would).

What kind of outdated piece of shit are you running that can't even render the up/down symbols?

he's blocking fonts