Why isn't it used more?
>>54319646
No idea really, v9/opus shit all over v8/vorbis and anything else out there.
>>54319646
html5
>>54319779
But it IS supported by html5. The only major stopping factor for wider adoption is lack of Apple support for the format. Even the fucking MS Edge is going to add support for this shit soon.
>>54319646
Isn't it because of lack of gpu accelerated decoding?
>>54319814
This. Chromium-based browsers use a ridiculous amount of CPU when playing Jewtube videos, because they force VP9 and no GPU out there can decode it. There's an extension called h264ify that forces it to use the sources Firefox gets, which drops CPU usage by around 90%.
>>54319646
Have you actually tried encoding something with it?
>takes forever and a day
>first try: end up with no keyframes, unseekable video
>second try, force keyframes every five seconds. file size is huge.
I am not the only one:
https://ffmpeg.zeranoe.com/forum/viewtopic.php?t=3032
https://github.com/fnordware/AdobeWebM/issues/8
Had to fall back to VP8 to make a usable webm.
Opus on the other hand seems to work nicely, but so does vorbis and as long as it's coupled with video any improvement in audio compression is all but insignificant anyway.
Meanwhile, h.264 encodes twice as fast with half the bitrate and no problems at all compared to VP8 with comparable quality. Feels good not to have software patents.
>>54320744
I don't understand this keyframe meme. The VP9 videos I can so far only use on x2 chan work just fine.
VP9 is also faster than VP8 in single-threading.
It kills battery on notebooks/phones/tablets.
>>54320744
Encoding VP9 on an i5-4690 with all 4 cores got me about 1 or 2 FPS at a fairly sensible quality setting.
>>54320954
Not everything is a meme. When you seek in a video (jump to some time point), the player needs to establish a base with the nearest keyframe before that point and fast-forwards from there. Without them, some players like the chromium-internal one stll manage to seek but take a long time because they have to start all the way from the beginning. VLC on the other hand just says "well fuck your shit I'm mainly a streaming player anyway" when it cannot find a keyframe and refuses to seek.
>>54321054
Yeah well one of your cores is more powerful than my whole poorfag cpu. But okay, I accept that encoding time is not really an issue then (or is it? what about energy usage on phones? h.264 is still an order of magnitude faster). My main concern is with the keyframes vs. file size thing anyway,
>>54320744
>takes forever and a day
It's slow yes but I can encode stuff at 3-4 fps on most content with speed 1 granted I have a very powerful and overclocked CPU. Speed 0 reduces the encoding speed by tenfold and gives little to no benefits in efficiency.
>first try: end up with no keyframes, unseekable video
>second try, force keyframes every five seconds. file size is huge.
Never had these issues and I've probably encoded up to a hundred little VP9 clips with the ffmpeg Windows builds provided by Zeranoe.
The encoder is slow but not too slow to use. I would definitely keep using it for videos posted here for example if it was supported. The efficiency compared to VP8 is quite huge once we go to low bitrates which is nice when you have a 3 Mb file size limit for webms like on 4chin. Opus is all around superior to Vorbis, especially at low bitrates.
thanks no
>>54319646
slow as mollusk to encode because unoptimized.
it's not any better than x264 10butt.
Was going to encode my torrents in VP9, but x265 seemed like it has more potential for future compatibility on mobile devices and consoles.
>>54323199
Does h265 even have an official file format?
4k blurays use h265
h265 won this time
>>54323883
I don't think online media will be using it for a while.
Not as long as MPEG-LA exists.
http://forum.doom9.org/showthread.php?p=1759920#post1759920
>Not using superior Nvidia GTX 950/960 and enjoy VP9 hardware decoding at very low CPU usage
PLEBS ALL OF YOU
4chan said no.
We are legion.
>>54322902
Interesting. I have the keyframe issue with the ffmpeg package for ubuntu 14.04. I triedffmpeg -i input.mp4 -c:v libvpx-vp9 -crf 10 -b:v 0 -g 120 -c:a libopus output.webm
This results in a huge file with the "-g 120" to force keyframes, and an okay but unseekable file otherwise.
Care to post your commandline?
>>54324220
>950/960
Sorry, I'm not poor.
>>54324404
Different ffmpeg enthusiast anon here. Why are you giving ffmpeg a video bit rate of 0?? And I could not find out what the -g option does. Care to share?
On a related note, I have never ever had problems with seeking or keyframes while using standard vp9 options, although I avoid the codec like the plague because
>1.2fps
So I did a test for you:$ffmpeg -version
ffmpeg version 3.0.1 Copyright (c) 2000-2016 the FFmpeg developers
and the command I used is:$ ffmpeg -i INPUT -c:v libvpx-vp9 -crf 10 -b:v 1M -c:a libopus -b:a 128K -ac 2 -threads 1 -y out.webm
The resulting video looks great and seeks well albeit slowly due to my weak CPU. I'd share it but neither 4chan nor webm upload sites accept it (maybe it's the opus audio?? EDIT: No, they refuse vp9).
Try re-encoding without the -g option or with a non-zero video bitrate, or both and report back.
>>54324404
I'm doing two pass encodes.
Highest quality encodes:
First pass:ffmpeg -i input.mkv -c:v libvpx-vp9 -pass 1 -b:v 3000K -threads 1 -speed 0 -tile-columns 0 -frame-parallel 0 -auto-alt-ref 1 -lag-in-frames 25 -g 9999 -aq-mode 0 -an -f webm D:\firstpass
On Linux you'll want to use a directory in your OS to store the temporary file for the first pass e.g. /dev/null instead of D:\firstpass
Second pass:ffmpeg -i output.mkv -c:v libvpx-vp9 -pass 2 -b:v 3000K -threads 1 -speed 0 -tile-columns 0 -frame-parallel 0 -auto-alt-ref 1 -lag-in-frames 25 -g 9999 -aq-mode 0 -c:a libopus -vbr on -compression_level 10 -b:a 128k -f webm output.webm
For much faster and still very good results just change the speed to 1 here. For test purposes I'm encoding a longer file with it right now the speed 0 setting. Currently getting 3 frames a second with it using the newest Zeranoe build and i7 4770k @ 4.3 GHz with HT turned off. Encoding on all cores, task manager showing about 40% CPU usage. Last time I used speed 0 I had HT turned on and the ffmpeg build probably had a different version of libvpx. HT seems not to work on some encoders and caused negative performance on VP8 for me.
For faster encodes I use these settings:ffmpeg -i input.mp4 -c:v libvpx-vp9 -pass 1 -b:v 3000K -threads 4 -speed 4 -tile-columns 2 -frame-parallel 1 -auto-alt-ref 1 -lag-in-frames 25 -an -f webm D:\firstpassffmpeg -i input.mp4 -c:v libvpx-vp9 -pass 2 -b:v 3000K -threads 4 -speed 1 -tile-columns 2 -frame-parallel 1 -auto-alt-ref 1 -lag-in-frames 25 -c:a libopus -b:a 128k -f webm output.webm
>>54319801
It's already in edge. I'm sad that I could watch webm if 4chan used VP9, but I can't because edge doesn't support VP8
>tfw two minutes of yuru yuri with full band stereo audio and 720p vp9 video in 2.4 mb
https://a.uguu.se/kwqkvo.webm
>>54325065
Looks like absolute shit though, so who cares?
>>54325122
It's far from the original quality yes. With any other encoder than h265 or vp9 that would look and sound like garbage at these bitrates. Garbage as in not watcable.
>>54324745
-g sets the maximum number of frames between keyframes, so 24 fps x 5 sec = 120.
Zero bitrate is vp9's "constant quality" mode, see
https://trac.ffmpeg.org/wiki/Encode/VP9
From what I understand this should be no different than constrained quality with b:v set really high.
Trying again now with your settings and then a two-pass encode. Might take a while.
Oh and alsoffmpeg -version
ffmpeg version N-79139-gde1a0d4 Copyright (c) 2000-2016 the FFmpeg developers
dude wat
>>54325271
this anime is unwatchable no matter the bitrate though
>>54325674
every anime is unwatchable, you silly manlets
>>54324831
different anon here... thanks
It isn't used more because the anime scene is retarded and thinks vp9 is just h265 but worse when it is actually better.
>>54325460
Finished testing. With the settings from this anon >>54324745 there are still no keyframes. The two-pass settings here >>54324831 give a better result with 3 keyframes in a 5 minute video. Still not a lot but at least the result is seekable with VLC.
>>54326301
They're literally about the same, minus the licensing agreements that the anime scene doesn't give two shits about. And HEVC is actually going to be used in 4K Bluray, which means it's not some meme like VP9 which is already being outdated by VP10 after Google saw that no one gave two shits about VP9.
https://arewecompressedyet.com/?r%5B%5D=x264-ntt-short-1b&r%5B%5D=x265_1.6_ntt-short-1&r%5B%5D=libvpx-vp9-1.4.0-ntt-short-1&s=ntt-short-1
