Welcome to Doom9's Forum, THE in-place to be for everyone interested in DVD conversion.

Before you start posting please read the forum rules. By posting to this forum you agree to abide by the rules.

 

Go Back   Doom9's Forum > Video Encoding > VP9 and AV1

Reply
 
Thread Tools Search this Thread Display Modes
Old 5th April 2018, 15:47   #601  |  Link
mzso
Registered User
 
Join Date: Oct 2009
Posts: 930
Quote:
Originally Posted by wiak View Post
av1 is less than 1%
So row-mt would improve it to 2%? What relevance does that have? So far it had zero performance optimizations. (As far as I know)
mzso is offline   Reply With Quote
Old 5th April 2018, 16:15   #602  |  Link
wiak
Registered User
 
Join Date: Jul 2003
Location: somewhere north
Posts: 260
Quote:
Originally Posted by mzso View Post
So row-mt would improve it to 2%? What relevance does that have? So far it had zero performance optimizations. (As far as I know)
hehe its so slow even time would be waiting for it to end before the universe does
__________________
Woah! Ninja?! http://nwgat.ninja/ (AV1 Overview)
"Not available in your region" has now been redefined as "Go Pirate, you filthy scum" Nwgat
wiak is offline   Reply With Quote
Old 5th April 2018, 20:30   #603  |  Link
mandarinka
Registered User
 
mandarinka's Avatar
 
Join Date: Jan 2007
Posts: 729
Quote:
Originally Posted by hajj_3 View Post
The EVE vp9 encoder claims to have better compression and offer faster encode times than x265.
Treat those claims like you would any marketing speak. It might be made by FFmpeg/open source guy, but it is still a vendor marketing pitch, he obviously only states things that make Eve look good.

After all, the thing is not publicly available, so nobody can really prove the claims.
mandarinka is offline   Reply With Quote
Old 5th April 2018, 23:25   #604  |  Link
hajj_3
Registered User
 
Join Date: Mar 2004
Posts: 1,120
Quote:
Originally Posted by mandarinka View Post
Treat those claims like you would any marketing speak. It might be made by FFmpeg/open source guy, but it is still a vendor marketing pitch, he obviously only states things that make Eve look good.

After all, the thing is not publicly available, so nobody can really prove the claims.
They seem to imply that google and netflix use the eve encoder.
hajj_3 is offline   Reply With Quote
Old 6th April 2018, 15:49   #605  |  Link
Djfe
Registered User
 
Join Date: May 2014
Posts: 24
might be, it's probably not that hard to improve vp9 for someone like him, and he worked at Google creating vp8 and vp9.

it certainly can be better, but that doesn't proof that it's always better (how he implies it), since the test samples can be constructed to get the statistics he wants.

"I only believe in statistics that I doctored myself"

I'm very certain, that it's better than libvpx in every way so it's likely that Google and especially Netflix (being big content providers) are interested in these gains.
but that doesn't make his statistics into a proof.
Djfe is offline   Reply With Quote
Old 7th April 2018, 21:10   #606  |  Link
mandarinka
Registered User
 
mandarinka's Avatar
 
Join Date: Jan 2007
Posts: 729
Yeah, it should definitely be better than libvpx, that's why Ronald Bultje (the author) wrote it after all! Not sure if Google uses it. At least for youtube they used to stick to libpvx and just didn't seem to care much about problems it has/had.

(Note that I didn't want to suggest the claims are outright dirty lies, just that thing the authors would claim tend to be one-sided, be it with On2, AV1, this case, x264/MCW...)
mandarinka is offline   Reply With Quote
Old 8th April 2018, 11:51   #607  |  Link
Mystery Keeper
Beyond Kawaii
 
Mystery Keeper's Avatar
 
Join Date: Feb 2008
Location: Russia
Posts: 724
libvps does have problems though.
I reported this critical bug a year ago, and they still haven't gotten to fixing it.
https://bugs.chromium.org/p/webm/iss...wner%20Summary
__________________
...desu!
Mystery Keeper is offline   Reply With Quote
Old 10th April 2018, 09:28   #608  |  Link
mzso
Registered User
 
Join Date: Oct 2009
Posts: 930
By the way. Is AVIF getting finalized in parralel with AV1?
Or is that something that will happen later, if at all?
mzso is offline   Reply With Quote
Old 10th April 2018, 15:44   #609  |  Link
bstrobl
Registered User
 
Join Date: Jun 2016
Posts: 55
Quote:
Originally Posted by mzso View Post
By the way. Is AVIF getting finalized in parralel with AV1?
Or is that something that will happen later, if at all?
I think they are standardising an AV1 still frame so that hardware decoders can be used to decode images in future. Container support in AVIF will come later since video is a priority.
bstrobl is offline   Reply With Quote
Old 10th April 2018, 15:45   #610  |  Link
bstrobl
Registered User
 
Join Date: Jun 2016
Posts: 55
New Xiph blog post:

https://people.xiph.org/~xiphmont/demo/av1/demo1.shtml
bstrobl is offline   Reply With Quote
Old 10th April 2018, 16:14   #611  |  Link
mzso
Registered User
 
Join Date: Oct 2009
Posts: 930
Quote:
Originally Posted by bstrobl View Post
I think they are standardising an AV1 still frame so that hardware decoders can be used to decode images in future. Container support in AVIF will come later since video is a priority.
The mention "animation" for AVIF. So they're going to do that with I frames?
This sounds remarkably counter-intuitive to me. Take only the I frames of a video codec, then make a half-assed video format ("animation") out of it.
mzso is offline   Reply With Quote
Old 10th April 2018, 16:58   #612  |  Link
bstrobl
Registered User
 
Join Date: Jun 2016
Posts: 55
Quote:
Originally Posted by mzso View Post
The mention "animation" for AVIF. So they're going to do that with I frames?
This sounds remarkably counter-intuitive to me. Take only the I frames of a video codec, then make a half-assed video format ("animation") out of it.
From what I can tell the still image header is optional and mainly designed to save a few bytes: https://aomedia-review.googlesource.com/c/aom/+/54781

I am going to take an educated guess and say that they will include Inter-frame compression for timed image sequences, just like HEIF/HEIC does, as it would be silly to waste storage not including it while reclaiming a few bytes in the header. It is possible that the logic for those components will be shifted to the P frames themselves to allow very basic decoders to deal with single frame images, as each frame is stored separately.
bstrobl is offline   Reply With Quote
Old 11th April 2018, 01:55   #613  |  Link
Clare
Registered User
 
Join Date: Apr 2016
Posts: 61
AV1 beats x264 and libvpx-vp9 in practical use case

https://code.facebook.com/posts/2538...ical-use-case/
Clare is offline   Reply With Quote
Old 11th April 2018, 04:27   #614  |  Link
foxyshadis
ангел смерти
 
foxyshadis's Avatar
 
Join Date: Nov 2004
Location: Lost
Posts: 9,558
Quote:
Originally Posted by bstrobl View Post
Whoa, awesome. Monty's posts are some of the best reading out there on the state of the art of AV compression, glad to see him back.

Quote:
Originally Posted by Clare View Post
AV1 beats x264 and libvpx-vp9 in practical use case

https://code.facebook.com/posts/2538...ical-use-case/
Interesting how they didn't even bother with x265. They've obviously rejected it for licensing reasons, even if they haven't ever said so. The encoding time increase for AV1 though... wow.

Last edited by foxyshadis; 11th April 2018 at 04:44.
foxyshadis is offline   Reply With Quote
Old 11th April 2018, 06:24   #615  |  Link
iwod
Registered User
 
Join Date: Apr 2002
Posts: 756
Quote:
Interesting how they didn't even bother with x265. They've obviously rejected it for licensing reasons, even if they haven't ever said so. The encoding time increase for AV1 though... wow.
I think somewhere along the line of HEVC licensing, one or all of the HEVC guys ( MPEG / HEVC Advance or Valos ) must have seriously pissed off these companies. I can literally smell and taste their hatred on HEVC, and vowed not to use it in their life time sort of attitude.
iwod is offline   Reply With Quote
Old 11th April 2018, 13:05   #616  |  Link
Shevach
Video compressionist
 
Join Date: Jun 2009
Location: Israel
Posts: 126
AV1 supports 1/8-pel motion precision (optional, specified by frame header parameter - allow_high_precision_mv). i wonder what's a gain in coding efficiency to exploit high MV precision? In HEVC and AVC the precision is 1/4-pel (for luma). In case of high-frequency video (HFR) or even for 60 fps 1/8-pel probably is redundant. Perhaps, 1/8-pel MV precision is beneficial for 4K resolution with 30 fps rate?
Unlike to HEVC development (where all was public and all discussions/contributions were located at jct-vc repository), some AV1's decisions are non-graspable to me. Who knows - under what circumstances 1/8-pel motion vector precision is beneficial vs. 1/4 pel one?
Shevach is offline   Reply With Quote
Old 11th April 2018, 13:12   #617  |  Link
Shevach
Video compressionist
 
Join Date: Jun 2009
Location: Israel
Posts: 126
Maybe for low-resolutions and low frequency sub-pixel movements achieve the 1/8 granularity?
Shevach is offline   Reply With Quote
Old 11th April 2018, 23:43   #618  |  Link
ianken
Guest
 
Posts: n/a
With a current FFMPEG build I get 0.2 fps on 1080p content. And that's a single stream.

To use this for next-day TV VOD it needs to be...faster. Or, maybe the ffmpeg integration is not fully baked? The CPU certainly is not loaded in any appreciable way. It's not even clocking up.
  Reply With Quote
Old 12th April 2018, 00:31   #619  |  Link
Tommy Carrot
Registered User
 
Tommy Carrot's Avatar
 
Join Date: Mar 2002
Posts: 863
The bitstream is not frozen, so you should use av1 only for testing... but currently it's too slow even for that. Once it is frozen, i expect the speed optimizations will come fairly quickly. Their stated goal is around a hundred times faster encoding speed at the end of the year. Hopefully with not significant quality drop.
Tommy Carrot is offline   Reply With Quote
Old 12th April 2018, 01:12   #620  |  Link
vidschlub
Registered User
 
Join Date: May 2016
Posts: 20
Quote:
Originally Posted by colinhunt View Post
Could you tell me where I can find this filter on x265? I went through its documentation and found nothing applicable.
You might need to set a +film flag or disable 'nograin' I honestly can't recall, I'm really a newbie at best, others here can help.

I've done a lot of side by side comparisons and if you don't have the side by side, you can trick yourself into thinking it's better, but a lot of detail is lost and washed out.

Someone here will know for sure.
vidschlub is offline   Reply With Quote
Reply

Thread Tools Search this Thread
Search this Thread:

Advanced Search
Display Modes

Posting Rules
You may not post new threads
You may not post replies
You may not post attachments
You may not edit your posts

BB code is On
Smilies are On
[IMG] code is On
HTML code is Off

Forum Jump


All times are GMT +1. The time now is 09:17.


Powered by vBulletin® Version 3.8.11
Copyright ©2000 - 2024, vBulletin Solutions Inc.