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 > New and alternative video codecs

Reply
 
Thread Tools Search this Thread Display Modes
Old 6th July 2018, 00:04   #741  |  Link
TD-Linux
Registered User
 
Join Date: Aug 2015
Posts: 32
Quote:
Originally Posted by Phanton_13 View Post
They are actually doing 16 arithmetic using 32 bit operations due that modern processor are faster unsing 32 bits operations that using 16bit operations, also varios presentations and documents indicates that dala range coder uses 15x16-> 31bit multiplications.
Since those slides were made, they shrank even further - they are only 7x9 multiplications. Lower latency for this multiply was critical for hardware throughput, so we shrank it as much as we could without significant compression losses. See https://aomedia.googlesource.com/aom...p/entdec.c#199

Quote:
Has AOMedia already started optimizing AV1 (especially its speed) after the bitstream freezed on the 25th of June?
Yup! You can always follow the latest development of libaom on Gerrit. Here's the most recent merged changes: https://aomedia-review.googlesource.com/q/status:merged
TD-Linux is offline   Reply With Quote
Old 6th July 2018, 00:19   #742  |  Link
MoSal
Registered User
 
Join Date: Jun 2013
Posts: 80
Had some time to do some tests with --cpu-used=4 and no tiles.

As the bitrate goes up, libaom continues to be superior compared to tuned x265 when it comes to I frames. For non-I frames, the story is different. Tuned x265 (see my previous comment) non-I frames look superior to libaom P frames which look terrible. Even when I give more bitrate to P frames (--max-intra-rate=700), they continue to look terrible.

The flexibility in the API and CLI options is quite lacking. The documentation is not great either (I had to check the source code to know what to pass to some options).

Again, this is with --cpu-used=4 which is probably analogous to veryfast in x264/x265. And it is needless to say that none of this is due to inherent limitations in the codec itself.
__________________
saldl: a command-line downloader optimized for speed and early preview.
MoSal is offline   Reply With Quote
Old 7th July 2018, 08:08   #743  |  Link
dapperdan
Registered User
 
Join Date: Aug 2009
Posts: 179
Quote:
Originally Posted by Mjpeg View Post
Great Chris Montgomery article on CDEF

https://hacks.mozilla.org/2018/06/av...cement-filter/
Interesting tidbit from Monty in the comments, sounds like Mozilla is working on its own AV1 encoder.

"Thats much more to do with the way the encoder is tuned, and youre seeing VPxs traditional tuning biases here. The VP encoders have always been rather merciless to mid-frequency content, and I think the current AOM encoder isnt yet using any sort of activity masking.

Compare AV1 to Daala on those images; the substantial difference there is difference in encoder tuning philosophy. Well be bringing that to the encoder were working on here at Moz."
dapperdan is offline   Reply With Quote
Old 7th July 2018, 08:45   #744  |  Link
Quikee
Registered User
 
Join Date: Jan 2006
Posts: 41
Quote:
Originally Posted by dapperdan View Post
Interesting tidbit from Monty in the comments, sounds like Mozilla is working on its own AV1 encoder.
Yup - rav1e - https://github.com/xiph/rav1e
Quikee is offline   Reply With Quote
Old 7th July 2018, 22:15   #745  |  Link
hajj_3
Registered User
 
Join Date: Mar 2004
Posts: 893
https://www.bbc.co.uk/rd/blog/2018-0...eg-and-aomedia
hajj_3 is offline   Reply With Quote
Old 9th July 2018, 06:28   #746  |  Link
foxyshadis
ангел смерти
 
foxyshadis's Avatar
 
Join Date: Nov 2004
Location: Lost
Posts: 9,413
Quote:
Originally Posted by Quikee View Post
Wow, I'm suddenly excited about AV1 again. Why is it that anything Monty does is like a kiss of gold to me?
__________________
There are four boxes to be used in defense of liberty: soap, ballot, jury, and ammo. Please use in that order.
foxyshadis is offline   Reply With Quote
Old 9th July 2018, 15:47   #747  |  Link
Barough
Registered User
 
Barough's Avatar
 
Join Date: Feb 2007
Location: Sweden
Posts: 322
AOM AV1 v1.0.0-82-gf77d93175
Built on July 09, 2018, GCC 7.3.0

Code:
https://aomedia.googlesource.com/aom

Last edited by Barough; 9th July 2018 at 17:48.
Barough is offline   Reply With Quote
Old 9th July 2018, 16:19   #748  |  Link
Zebulon84
Registered User
 
Join Date: Apr 2015
Posts: 12
Quote:
Originally Posted by Barough View Post
AOM AV1 v1.0.0-82-gf77d93175
Built on April 19, 2018, GCC 7.3.0
I guess the "April 19, 2018" is just a copy-paste fromthat previous message. Files in the archive are from today, July 9, 2018.
Zebulon84 is offline   Reply With Quote
Old 9th July 2018, 17:50   #749  |  Link
Barough
Registered User
 
Barough's Avatar
 
Join Date: Feb 2007
Location: Sweden
Posts: 322
Quote:
Originally Posted by Zebulon84 View Post
I guess the "April 19, 2018" is just a copy-paste fromthat previous message. Files in the archive are from today, July 9, 2018.
Thnx for the heads up. As u said...... it's copy/paste from the old post. Fixed now.

Last edited by Barough; 9th July 2018 at 18:05.
Barough is offline   Reply With Quote
Old 9th July 2018, 19:27   #750  |  Link
benwaggoner
Moderator
 
Join Date: Jan 2006
Location: Portland, OR
Posts: 2,874
Quote:
Originally Posted by Quikee View Post
The VPx series was quite unusual in that the bitstream proponent also made the primary encoder. For all the MPEG codecs, MPEG didn't worry much about the speed of the reference encoder because only 3rd party implementations are used in the market.

It's a good sign of the health of AV1 that we are seeing a variety of different encoders, both open-source and proprietary, being worked on. VPx never had enough interest to get that effort.

Competition drives tons of implementation innovations in encoding. That's why we're still seeing significant improvements in MPEG-2 after all these years.
__________________
Ben Waggoner
Principal Video Specialist, Amazon Prime Video

My Compression Book
benwaggoner is offline   Reply With Quote
Old 9th July 2018, 22:04   #751  |  Link
utack
Registered User
 
Join Date: Apr 2018
Posts: 38
Quote:
Originally Posted by benwaggoner View Post
VPx never had enough interest to get that effort.
That is not entirely true, there is Eve, which Netflix used for a while
utack is offline   Reply With Quote
Old 10th July 2018, 02:04   #752  |  Link
soresu
Registered User
 
Join Date: May 2005
Location: Swansea, Wales, UK
Posts: 86
I was under the impression that Ronald Bultje from the libvpx team built Eve, so while technically separate its practically incestuous in creation
soresu is offline   Reply With Quote
Old 10th July 2018, 07:21   #753  |  Link
iwod
Registered User
 
Join Date: Apr 2002
Posts: 749
Quote:
Originally Posted by benwaggoner View Post
Competition drives tons of implementation innovations in encoding. That's why we're still seeing significant improvements in MPEG-2 after all these years.
I think I have read this statement sometime ago and may have asked a similar question before. What are the use case of MPEG-2 today? From high end bitrate to low end I cant think of a single use case where it would be valuable.
iwod is offline   Reply With Quote
Old 10th July 2018, 08:29   #754  |  Link
Blue_MiSfit
Derek Prestegard IRL
 
Blue_MiSfit's Avatar
 
Join Date: Nov 2003
Location: Los Angeles
Posts: 5,504
Quote:
Originally Posted by iwod View Post
I think I have read this statement sometime ago and may have asked a similar question before. What are the use case of MPEG-2 today? From high end bitrate to low end I cant think of a single use case where it would be valuable.
So many.

- Terrestrial broadcast in the US using ATSC is 100% MPEG-2 for now.
- Legacy cable / satellite networks with tons of existing decoders / professional IRDs in the field
- Legacy broadcast contribution, especially at high bitrates for backup.
- Legacy cable VOD (still pushing 15 Mbps 1080i MPEG-2 in most cases)
- Broadcast playout - quite heavily using 50 Mbps XDCAM HD422, sometimes even lower quality.

TBH, for anything new that's for distribution, yeah you wouldn't use MPEG-2 most likely.

However, legacy stuff has a habit of staying around forever.

Plenty of current premium satellite TV networks use MPEG-2 as their house format because surrounding standards like XDCAM HD422 in MXF have robust support for in-band metadata like captioning, timecode, AFD, etc, and also have broad support from playout server vendors, NLE / post production tools, and pro transcoder tools. The other benefit is that MPEG-2 is quite lightweight to decode these days, so a video server can be dense and cost effective and still perform perfect frame accurate seeking and smooth playback, saving CPU cycles for graphics etc.

Quality is definitely "good enough" (especially considering transmission encoding typically being 6-12 Mbps real-time encoded CBR H.264 with small GOPs and tight buffers).

More modern alternatives like J2K (a-la AS-02 style MXF) and AVC Intra do have benefits in certain cases, especially where very high quality is desired, but they typically come with additional cost in terms of processing power, software licensing, and storage capacity. Higher quality formats tend to live in acquisition and post production. When you deliver to playout, 1080i XDCAM HD422 is probably the most common standard - at least in the US.

The broadcast industry as a whole tends to cling to the melting icebergs of trust for as long as possible, and TBH this is for good reason. Any change that introduces potential risk is an extremely tough sell when you have 4 or 5 nines of uptime required in your SLA. It only takes a few minutes of downtime per year to start feeling the pain.

I've heard of some improvements in MPEG-2 especially for H.264 -> MPEG-2 transcoders to glue new channels into legacy infrastructure, but TBH this was years ago. I haven't heard of anything major recently. Ben, care to share?

Last edited by Blue_MiSfit; 10th July 2018 at 08:35.
Blue_MiSfit is offline   Reply With Quote
Old 10th July 2018, 21:49   #755  |  Link
benwaggoner
Moderator
 
Join Date: Jan 2006
Location: Portland, OR
Posts: 2,874
Quote:
Originally Posted by Blue_MiSfit View Post
I've heard of some improvements in MPEG-2 especially for H.264 -> MPEG-2 transcoders to glue new channels into legacy infrastructure, but TBH this was years ago. I haven't heard of anything major recently. Ben, care to share?
Elemental's encoder has delivered >20% bitrate reduction for statmuxed MPEG-2 over the last couple of years.

Granted statmuxed MPEG-2 is kind of a special case, except that it probably accounts for the majority of MPEG-2 eyeball hours these days.

(for those blissfully ignorant of channel-based broadcast world, statmuxing is when multiple video streams are encoded in parallel to fit within a given amount of total bandwidth. Basically inter-stream VBR).
__________________
Ben Waggoner
Principal Video Specialist, Amazon Prime Video

My Compression Book
benwaggoner is offline   Reply With Quote
Old 12th July 2018, 08:20   #756  |  Link
hajj_3
Registered User
 
Join Date: Mar 2004
Posts: 893
https://medium.com/@luc.trudeau/av1-...s-6871007ad99e
hajj_3 is offline   Reply With Quote
Old 12th July 2018, 11:21   #757  |  Link
paul97
Registered User
 
Join Date: Mar 2018
Posts: 3
When there''ll be a good and official encoder for AV1. (for example in Handbrake or Youtube. We need to wait for Mozilla''s Rust Encoder or they immediately will release AOMedia encoder?
paul97 is offline   Reply With Quote
Old 12th July 2018, 20:05   #758  |  Link
LigH
German doom9/Gleitz SuMo
 
LigH's Avatar
 
Join Date: Oct 2001
Location: Germany, rural Altmark
Posts: 5,843
Official? Well, aomenc is already available as codec for ffmpeg; and HandBrake uses an ffmpeg core, you just need to build a recent version or hope for one to be released.

Whether it is "good" ... your demands, you opinion.
__________________

New German Gleitz board
MediaFire: x264 | x265 | VPx | AOM | Xvid
LigH is offline   Reply With Quote
Old 12th July 2018, 20:06   #759  |  Link
benwaggoner
Moderator
 
Join Date: Jan 2006
Location: Portland, OR
Posts: 2,874
Quote:
Originally Posted by paul97 View Post
When there''ll be a good and official encoder for AV1. (for example in Handbrake or Youtube. We need to wait for Mozilla''s Rust Encoder or they immediately will release AOMedia encoder?
What do you consider a "good official" encoder for H.264 and HEVC? Generally for the MPEG codecs there is a reference encoder which is way too slow for production use, and then a variety of vendors making their own implementations, some open-source, some proprietary.

I wouldn't expect to have everyone using one "official" encoder like libvpx with AV1 if AV1 gets significantly broad market adoption. VPx had mainly a very small number of very large companies using it.
__________________
Ben Waggoner
Principal Video Specialist, Amazon Prime Video

My Compression Book
benwaggoner is offline   Reply With Quote
Old 13th July 2018, 03:11   #760  |  Link
olduser217
Registered User
 
Join Date: Jun 2015
Posts: 17
Quote:
Originally Posted by Quikee View Post

From encoding strategy or algorithm point of view, are these AV1 encoders developed by Mozilla or EVE have any difference compare to the one from AOM? Or just mainly encoding speed up?

Last edited by olduser217; 13th July 2018 at 03:23.
olduser217 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 12:13.


Powered by vBulletin® Version 3.8.11
Copyright ©2000 - 2019, vBulletin Solutions Inc.