What's the meta for encoding VP9 stuff for Any Forums?
I keep seeing conflicting arguments being made for bit depth, color spaces, and bitrate control, all with a dose of partisanship.
What's the meta for encoding VP9 stuff for Any Forums?
Luis Gray
Other urls found in this thread:
chromium.googlesource.com
argorar.github.io
twitter.com
Carter Wright
They work fine for me. My script just werks with a couple changes.
Evan Jenkins
Can't say the same about phoneposters. They are definitely screwed.
Jordan Ramirez
chromium.googlesource.com
Meta is using the latest libvpx, many tools are old and haven't been updated for years
Austin Thompson
I've never cared about them.
John Hill
AFAIK single pass bitrate control is still broken so you still need to perform a 2-pass, even for CRF mode. This freetard codec still requires you to manually enable lag-in-frame and auto-alt-ref to get comparable compression efficiency to SW HEVC, a problem carried forward from VP8, so it's still a CLI dependent codec if you want good quality. The good news is you can finally perform 2 pass (ABR at least) on handbrake so you can still achieve ~80% max compression efficiency of VP9 through that alone. It's slow as balls to encode so if you have 2500K level CPU performance or worse it's better if you stick with VP8 since the slowest encode of that will give you around the same quality as a quick realtime VP9 encode.
Substantial upgrade but it's still not a professional video codec like hevc which has 100+ FPS 1080p gpu encoding comparable to software x264 slow.
Gabriel Howard
Some neglected feature is the full range flag. It won't look right on some phones, but at least it plays. It can preserve the dynamic range of raw RGB and bring it on par with JPEG.
This can be used on a series of PNG's.
-filter:v "zscale=pin=bt709:min=gbr:tin=iec61966-2-1:rin=pc:p=bt709:m=bt709:t=iec61966-2-1:r=pc,format=yuv420p" -color_primaries bt709 -colorspace bt709 -color_trc iec61966-2-1 -color_range pc
Those last parameters may be redundant, as ffmpeg can add them automatically based on input.
Bentley Adams
Works for me. Is it an iOS thing I'm too Android to understand?
Hudson Kelly
phoneposter here, Kuroba EX didn't even need an update to support vp9 and the browser could already read vp9 aeons ago
Levi Perry
I have a machine encoding X265 HEVC on an i5 4590. That machine can encode a 20 minute episode faster than any machine I have can encode a VP9 2 minute webm. VP9 is pure ass for encoding speed.
Sure it's not the correct thing to encode video on but it is a job for a box that would otherwise be thrown out.
James Walker
the meta is to not be autistic
ffmpeg -i -crf -vf scale=-1:720 -an output.webm
Just change crf until you get the filesize you need (lower = higher quality and default is 32) as well as whatever resolution you need.
Gabriel Collins
posters here are retarded, I can see the new webms on iPhone just fine
Brody Reyes
who gives a shit just -an -c:v vp9 -b:v 2M and post your stupid webm
Ryder Edwards
Just use WebM for Lazys.
argorar.github.io
It optimises all the ffmpeg arguments for you to squeeze out quality, and just produced something that works properly on Any Forums.
Don't use the unmaintained WebM For Retards.
William Murphy
>Only works on Windows
Luke Young
That's why it was originally WebM for Retards.
But it is just convenient because it has handy trimming, cropping, resizing options and more.
Jayden Howard
That's cool until you encode a video with a color mode that browsers won't accept.
David Peterson
These opinions are not really conflicting.
They represent different points on the scale of “slow encoding for the higher quality” vs. “fast and simple encoding” where the latter still provides much better quality than VP8.
Here and are the examples of a simpler approach.
They would perfectly suffice for short video quotes that also do not depict much motion.
They won't suffice if you need a whole minute of a very dynamic fullHD in your 6 megabytes (such as for example), and then you'd use 2-pass, you'd kill all the unnecessary metadata (“-map_metadata -1 -map_chapters -1”), you'd analyze more frames (“-lag-in-frames 25”), you'd build better references in your video stream (“-auto-alt-ref 1 -arnr-maxframes 12 -arnr-strength 4 -enable-tpl 1”), etc.
Then you'll also notice how slow the encoding becomes, and you'd parallelize (for example, “-tile-columns 2 -threads 8” on eight cores or on four double-threaded cores).
Hence all the additional parameters in command lines.
Andrew Sanders
only thing I really added on top of `-c:v vp9` was `pix_fmt yuv420p` because my source videos are yuv444p
and while hiroshimoot enabled yuv444p I have had a few people (macfags?) quote my posts to tell me the webm is broken
Daniel Butler
not my problem, use a better browser
William Adams
Thanks, googling those options brought up a nice summary.
Ffmpeg's own doc isn't great.
Ethan Bennett
this webm is broken for me. only one in the thread.
Caleb Smith
First I've heard of that. Works fine on my machine.
Hudson Gomez
try 8 bit instead of 10?
Noah Reed
ffmpeg noob here. What is the distinction between -b:v and -crf flags?
Jose Richardson
I don't believe I encoded it in 10 bit.
This is the encode command.
ffmpeg -loglevel info -i "$sourcefile" -c:v libvpx-vp9 -ss $starttime -to $endtime -vf scale=$resolution:-1$subs$transpose -sn -b:v $size*1024*1024*8/$time*0.975 -quality good -r $FPS -threads $core -slices 8 -auto-alt-ref 1 -lag-in-frames 25 $audio -tile-columns 1 -row-mt 1 -pass 2 temp.webm #encode pass