HEVC support? (new standard from MPEG; 40-50% > than H264)

Archive of historical feature requests.
Please use the GitHub link above to report issues.
Forum rules
*******************************
Please be aware we are now using GitHub for issue tracking and feature requests.
- This section of the forum is now closed to new topics.

*******************************
hippo_powa
Posts: 8
Joined: Thu Aug 16, 2012 1:57 am

HEVC support? (new standard from MPEG; 40-50% > than H264)

Post by hippo_powa » Thu Aug 16, 2012 2:06 am

Dear lovely developers of Handbrake,

The MPEG dudes have just announced today that the draft standard of HEVC (sometimes referred to as H265) is out.

In short: twice the quality at the same bitrate, or the same quality at half the bitrate, relative to H264, plus support for some new resolutions.

Some linkys for the curious

http://www.bit-tech.net/news/bits/2012/ ... peg-hvec/1
http://www.engadget.com/2012/08/15/mpeg ... -standard/
http://hevc.hhi.fraunhofer.de/
http://en.wikipedia.org/wiki/High_Effic ... deo_Coding
http://www.tekbuz.com/qualcomm-compares ... emo/180099

A free halving of space requirements is something that I imagine everyone would be very happy about, from mobile phone users to people backing up their DVD collection.

On what timeframe do you think Handbrake might support this codec (or something based on it?)

All the best and thanks again for your amazing work

Hippo hippo

saintdev
Regular User
Posts: 146
Joined: Wed Dec 20, 2006 4:17 am

Re: HEVC support? (new standard from MPEG; 40-50% > than H26

Post by saintdev » Thu Aug 16, 2012 7:54 am

You need to realize that most of the HandBrake team are not codec developers. They simply make use of libraries written by other people. So on this note, they have very little say over if/when HEVC will be added. Basically it comes down to "When someone else writes the code for it".

You also need to take into consideration that there is a lot of marketing involved here. There are a lot of companies that have patents involved and would like this standard to be adopted by everyone so they can sit back and make money. So when they say "50%" or "half" that is going to be under ideal conditions. Realistically you are more likely to see around 30% improvement, at best.

It's not really just as simple as "HandBrake adding support for this codec". You still need decoders to be written, then optimized. So you can, you know, actually watch the videos you have encoded. Hardware decoders, for low power devices, need designed, tested, modified, tested, modified, tested..... An open source encoder needs to be written, then optimized, then optimized more. None of these can be done until the final draft is released. This doesn't include the time required for people to adopt the new standard, which history has shown will take a while, even after you have the decoders and encoders available.

TedJ
Veteran User
Posts: 5388
Joined: Wed Feb 20, 2008 11:25 pm

Re: HEVC support? (new standard from MPEG; 40-50% > than H26

Post by TedJ » Thu Aug 16, 2012 12:18 pm

All in all, it's probably a waste of effort trying to implement this stop-gap codec, we should move straight onto developing support for H.268... ;)

tlindgren
Bright Spark User
Posts: 237
Joined: Sun May 03, 2009 2:14 pm

Re: HEVC support? (new standard from MPEG; 40-50% > than H26

Post by tlindgren » Thu Aug 16, 2012 3:50 pm

hippo_powa wrote:The MPEG dudes have just announced today that the draft standard of HEVC (sometimes referred to as H265) is out.

In short: twice the quality at the same bitrate, or the same quality at half the bitrate, relative to H264, plus support for some new resolutions.
You've been looking at press releases and taking them at face value, that's almost always a bad idea... Especially so with video standards.

Even if we assume nothing changes between draft and final standard, my personal best guess is that somewhere in 2014 we may see the first commercial h.265 encoder with better quality at the same bitrate than x264 (that HandBrake uses) encoding at a usable speed. My understanding is that what they have now is way beyond even "glacially slow" and I also suspect they're very careful with what they show (suggesting that even with insanely large clusters and accepting painfully slow encoding it often doesn't do "well enough").

And I suspect it's a fairly good chance the specification do change significantly from draft to final version which will likely push back both hardware and software, if this happens my best guess the crossover may be somewhere during 2015.

There's also the decoding side to look at, I expect that few handheld devices currently being sold will be able to handle this at 1080p without hardware assist which doesn't exist before 2013 at the earliest! I wouldn't be surprised if we're into 2015 or even 2016 before hardware decoding is common enough that it's really useful as a video encoding standard.

h.264 hardware decoding got a big head-start thanks to Blu-ray, yet we need to remember that the final draft (not "first usable draft" like this appears to be) was in May 2003... It was many, many years before good hardware decoders were common, but 9! years later there's still current hardware being sold which has issues with High Profile, specific features and/or can't handle decent bitrates!

With regards to a non-commercial implementation reaching the same point (beating x264) it's a bit harder to guess, if the x264 team goes after it in a large scale my guess is that there's a chance it could happen in late 2013 though even if all star aligns 2014 or later seems likely. And not all stars aligns we shouldn't be surprised if it'll be a while after the first good commercial implementations.

hippo_powa
Posts: 8
Joined: Thu Aug 16, 2012 1:57 am

Re: HEVC support? (new standard from MPEG; 40-50% > than H26

Post by hippo_powa » Thu Aug 16, 2012 10:21 pm

The only useful comment posted so far (in relation to the question being asked) is the comment that 'it'll be ready when someone else does it'. I don't know if that person speaks for the whole handbrake team, but it's at least useful as a starting point.

Rather than posting personal speculation and wild theories, can we stick to facts?

Here are some facts.

"The main stated goal of the HEVC development is “substantial” bitrate reduction relative to AVC High Profile. The general target of the group (although informal) is to provide about 50% compression improvements over AVC. A second goal is to serve a wide range of applications. To do this, HEVC defines support for resolutions from QVGA (320x240) to 8K (7680x4320). HEVC also targets two modes, a “low complexity” mode, which is meant to have a small decrease in complexity (especially for decoders) compared to AVC, and a “high efficiency” mode, which will contain more complex coding tools, achieve better compression, but come at higher processing costs. Finally, HEVC defines a “low delay” configuration capable of low latency operation. In general, HEVC decoders are expected to have 2-3 times the computational complexity of AVC decoders, and HEVC encoders are expected to have up to 10 times the computational complexity of AVC encoders."
http://www.sencore.com/company/blog/eme ... s-part-two

*up to* being the key phrase; there are two modes of encoding. Low-complexity, which has a similar run-time to H264 encoding but 25% better filesizes, and high-complexity, which takes 10x as long as H264 (or 100x as long as MPEG-2), but achieves 50% improvement with no drop in subjective quality. Glacially slow? Are you serious?

As for 'you've been looking at press releases'? No, I've been reading academic papers with benchmarks, and standards organisation releases. HEVC has thousands of people behind it from dozens of organisations, and many years of work. These people include 'non-corporate' participants such as academic researchers, the BBC and so on. The tests and benchmarks they run are standardised and were agreed upon at the start of development. There are already multiple open source implementations already out there, which a simple Google search will reveal.

http://hevc.kw.bbc.co.uk/git/w/jctvc-tmuc.git
http://code.google.com/p/x265/
https://hevc.hhi.fraunhofer.de/svn/svn_ ... gs/HM-1.0/ (reference decoder)

"The JCT-VC has published a software reference implementation of the proposed HEVC standard, called the “HEVC Test Model” or HM. The latest version is HM-7.0 (based on version 7.0 of the HEVC standard, which was proposed at the May 2012 meeting). This implementation is open-source and includes both a decoder and an encoder application. HHI hosts the subversion repository for the code and BBC hosts the issue tracker for the code. HHI also published a software reference manual and software development guidelines for the HM."
http://www.sencore.com/company/blog/eme ... s-part-two

There are already software and hardware products out there relating to h265/hevc, in preparation or complete.

http://www.solveigmm.com/en/products/zond/
http://hevcvisa.codecian.com/

Talk of a 2014 or later timescale is ignorant beyond imagination. The code's out there already. It works. Even hardware implementations already exist. And in terms of underlying design it's hardly a world away from existing codecs such as H264/x264. The almost-final reference has achieved the design goals.

In terms of efficiency goals, they did it!

"N12475 is the Report on preliminary subjective testing of HEVC compression capability which can be found here. It shows impressive results as reported elsewhere, e.g., here. In particular, > 50% bitrate reduction, 67% in class B (HDTV), 49% in class C (WVGA) => mission accomplished! Currently, HEVC is between ballots and FDIS/IS is expected around Jan-Apr 2013."
http://multimediacommunication.blogspot ... eting.html

Consider also, we are talking about a standard which is an evolutionary improvement upon H264 rather than revolutionary. Look how quickly 80211.n caught on, and that required a change of physical hardware , firmware and even network reconfiguration; yet enterprises were adopting 802.11n in practice before it was finalised officially (http://en.wikipedia.org/wiki/IEEE_802.11#802.11n).

In contrast, much of HEVC can be achieved in software and via the use of reprogrammable GPUs (openCL). Demands on CPU/GPU are relatively low in comparison to the quantum leap between MPEG-2 and H264. Literally every new computer sold has a built in multi-processor general purpose maths accelerator with exactly the right kinds of functions built in to support things like HEVC.

And what is the impetus to drive adoption that will make this be a reality for us all before e.g. mid-2013?

"Internet video was 40 percent of consumer Internet traffic in 2010 and will reach 50 percent by year-end 2012."
"Every second, 1 million minutes of video content will cross the network in 2015."

http://www.cisco.com/en/US/solutions/co ... Paper.html

Halving bitrate halves costs - of transmission, server storage, client storage. That's the kind of thing companies will buy into *very* easily. Video is the bulk of the world's data. We can effectively make the entire internet twice as fast, the world's storage twice as vast, double the number of channels on satellite or cable (or double their quality), by adopting this standard. How can anyone imagine the world is going to sit on its butt till '2014 or later' when that kind of opportunity is sitting there for the media and network companies of the world? This codec is the kind of thing that affects world GDP.

As for "Way beyond glacially slow". Are you joking? The whole point of the exercise was to develop a standard that offered similar encode times to H264 while producing a noticeably smaller file, or greater encode times with larger compression. Here's a factual example:

http://iphome.hhi.de/schierl/assets/hevc_icassp2012.pdf

12-core intel producing 1920x1080 at 51fps? Now what does that mean for someone encoding a DVD on a 4-8x core desktop?

Approximately: (1920*1080/12) / (720*576/8) * 51 = 170fps encoding of DVD on an 8-core desktop. This should be adjusted down slightly since Xeon chips run faster than regular ones - perhaps 130-150fps.

Sound impossible? Then consider that the whole purpose of the R&D was to develop an encoder with similar encode times to H264 and 25% file size improvement, making use of smarter tricks to achieve the compression (in the low complexity mode), or achieving twice the file reduction (50% vs 25%) with approx 10x the encode time of H264, in high complexity mode.

Decode isn't hard either. The whole point of the exercise was to design something with a light decode process based on similar mathematical functions to H264. Amazing as this might seem, all the technology companies of the world have actually noticed that people use digital cameras and mobile phones and tablet PCs.

"David Hopkins, director of product marketing, MPEG-4 encoding solutions group, Motorola, told CSI on the side of the company’s press and customer event in Stockholm that the complexity of HEVC - it’s about 100 times more complex than MPEG-2 to encode but only three to five times as complex to decode – means the technology will first appear in mobile devices, potentially before the end of 2013."
http://www.csimagazine.com/csi/OTT-will ... torola.php

The myth about 'glacially slow' encoders originated from one phase of development where effectively a competition was being held to see who could crunch the data the smallest (teams were allowed two entries: 'high complexity' and 'low complexity'. Unsurprisingly, some teams (e.g. the BBC team) decided to cheat a bit by effectively testing literally every possible compressed representation using an exhaustive search for their high complexity entry, to see which representation achieved the optimal compression. Cunning. It's also entirely irrelevant to real world use.

Read this thread from early 2010 to see the origination of the 'HEVC is glacially slow' myth:
http://x264dev.multimedia.cx/archives/360
in particular the comments by Hurumi, the BBC/Samsung development lead.

Some more interesting links with data and benchmarks:
https://dl.dropbox.com/u/1346434/w12475.zip
http://www.vcodex.com/h265.html

Now, with all that said. Is there anyone here who has actually been involved in the HEVC development process who can make a reasonable fact-based estimate of when we might see HEVC (or something closely related) show up in a form that can be easily incorporated into Handbrake?

Please avoid replying if you have no familiarity with HEVC or HEVC development. Thanks.

randomreuben
Veteran User
Posts: 468
Joined: Mon Nov 02, 2009 2:18 pm

Re: HEVC support? (new standard from MPEG; 40-50% > than H26

Post by randomreuben » Fri Aug 17, 2012 4:00 am

With hard drive space being as cheap as it is, and that nearly every device now decodes x264, why would you want to switch to another codec? x264 is like MP3 technology now. It's good enough for a lot of people and it's ubiquitous. Consider AAC vs MP3. Sure, AAC is SO much more awesome and gives its users nerdgasms, but a 256 kbps MP3 file is transparent to the end user (except for the three people who are blessed/cursed with ears that require this. http://www.positive-feedback.com/Issue32/anjou.htm) No matter how awesome AAC is, no one in their right mind would ever think of dropping MP3 support in their hardware players because a high bitrate MP3 file is indistinguishable from a high bitrate AAC file.

So going back to the HEVC, I'm sure this new codec will heal the sick and cause the lame to walk and the blind to see, but everyone's really enjoying using the x264 encoder, so who cares about some new codec that will do a little bit more for size?

You're really at the wrong place completely if you're looking for HEVC support to be added here immediately. For a start, there will need to be some GPL (or some such licenced) encoder available before it is added to HandBrake. Until that happens, you're out of luck for HEVC support.

saintdev
Regular User
Posts: 146
Joined: Wed Dec 20, 2006 4:17 am

Re: HEVC support? (new standard from MPEG; 40-50% > than H26

Post by saintdev » Fri Aug 17, 2012 5:25 am

hippo_powa wrote:Here are some facts.

"The main stated goal of the HEVC development is “substantial” bitrate reduction relative to AVC High Profile. The general target of the group (although informal) is to provide about 50% compression improvements over AVC. A second goal is to serve a wide range of applications. To do this, HEVC defines support for resolutions from QVGA (320x240) to 8K (7680x4320). HEVC also targets two modes, a “low complexity” mode, which is meant to have a small decrease in complexity (especially for decoders) compared to AVC, and a “high efficiency” mode, which will contain more complex coding tools, achieve better compression, but come at higher processing costs. Finally, HEVC defines a “low delay” configuration capable of low latency operation. In general, HEVC decoders are expected to have 2-3 times the computational complexity of AVC decoders, and HEVC encoders are expected to have up to 10 times the computational complexity of AVC encoders."
http://www.sencore.com/company/blog/eme ... s-part-two

*up to* being the key phrase; there are two modes of encoding. Low-complexity, which has a similar run-time to H264 encoding but 25% better filesizes, and high-complexity, which takes 10x as long as H264 (or 100x as long as MPEG-2), but achieves 50% improvement with no drop in subjective quality. Glacially slow?
Yes, unoptimized software is slow, especially in the video codec world. Modern hardware can barely encode 1080p in realtime using one of the most heavily optimized pieces of software out there (x264). 100x slower than this is poor, they are most likely comparing to JM, which is known to be about 100x slower than x264.
As for 'you've been looking at press releases'? No, I've been reading academic papers with benchmarks, and standards organisation releases. HEVC has thousands of people behind it from dozens of organisations, and many years of work. These people include 'non-corporate' participants such as academic researchers, the BBC and so on. The tests and benchmarks they run are standardised and were agreed upon at the start of development.
Academic papers are notourisly as bad as press releases.
There are already multiple open source implementations already out there, which a simple Google search will reveal.

http://hevc.kw.bbc.co.uk/git/w/jctvc-tmuc.git
http://code.google.com/p/x265/
https://hevc.hhi.fraunhofer.de/svn/svn_ ... gs/HM-1.0/ (reference decoder)

"The JCT-VC has published a software reference implementation of the proposed HEVC standard, called the “HEVC Test Model” or HM. The latest version is HM-7.0 (based on version 7.0 of the HEVC standard, which was proposed at the May 2012 meeting). This implementation is open-source and includes both a decoder and an encoder application. HHI hosts the subversion repository for the code and BBC hosts the issue tracker for the code. HHI also published a software reference manual and software development guidelines for the HM."
http://www.sencore.com/company/blog/eme ... s-part-two
First off the "x265" you link to is nothing more than a few headers (mostly copied from x264), and a few C++ files with nothing of substance in them. Also, so you don't get the wrong idea, it's author has no relation to any of the x264 development team.

HM on the other hand, is the reference software, just like JM was before it for AVC. While it's true, both are open souce, their license conflicts with most any other open source license (ie: anything GPL or GPL-compatible). This prevents them from being linked to or distributed by any open source project (such as HandBrake). Reference software has one design goal, to accurately implement the spec (It doesn't always accomplish this goal, but that is beside the point). That means that there is no work put into optimization. JM, even 10 years later on modern hardware, is still unusably slow. It's really only useful for verification purposes. HM is in the same boat.
There are already software and hardware products out there relating to h265/hevc, in preparation or complete.

http://www.solveigmm.com/en/products/zond/
http://hevcvisa.codecian.com/

Talk of a 2014 or later timescale is ignorant beyond imagination. The code's out there already. It works. Even hardware implementations already exist. And in terms of underlying design it's hardly a world away from existing codecs such as H264/x264. The almost-final reference has achieved the design goals.
Please refrain from calling people names. Let's keep this civil.

I still have yet to see a single hardware decoder, or a mention of a hardware decoder. Qualcomm (the video you youself linked to earlier) has stated they are waiting for the final draft before they start working on a hardware decoder. No chip manufacturer is stupid enough to commit silicon to a first draft that is going to go through many revisions before it's finalized.
As for the two bitstream analyzers you link to, those are a special case. What they do is, take the reference software modify it slightly, and put together a GUI. These are not independent implementations.
Consider also, we are talking about a standard which is an evolutionary improvement upon H264 rather than revolutionary.
MPEG-2 was an evolutionary improvement on MPEG-1. MPEG-4 ASP was an evolutionary improvement on MPEG-2. MPEG-4 AVC was an evolutionary improvement on MPEG-4 ASP. This means nothing. Look at how long it took each of those to become widely adopted.
In contrast, much of HEVC can be achieved in software and via the use of reprogrammable GPUs (openCL).
OpenCL is mostly useless in the relm of video codecs. GPUs have dedicated silicon just for video decoding, that is not reprogrammable. Have you noticed how every GPGPU video encoder has horrendous quality, this is because they have to sacrifice a lot of decisions to fit the GPU architecture at the cost of quality improvement.
Literally every new computer sold has a built in multi-processor general purpose maths accelerator with exactly the right kinds of functions built in to support things like HEVC.
Somebody still needs to write that software, it doesn't just magically use something because it happens to be there.
Halving bitrate halves costs - of transmission, server storage, client storage. That's the kind of thing companies will buy into *very* easily. Video is the bulk of the world's data. We can effectively make the entire internet twice as fast, the world's storage twice as vast, double the number of channels on satellite or cable (or double their quality), by adopting this standard. How can anyone imagine the world is going to sit on its butt till '2014 or later' when that kind of opportunity is sitting there for the media and network companies of the world?
If this were the sole requirement for adoption, a lot of things would be different in this world. US cable and terrestrial broadcasts are still MPEG-2, this is true of most of the world, despite AVC being a viable alternative at the time these standards were developed. Most Blu-rays are still released in MPEG-2.
Now, with all that said. Is there anyone here who has actually been involved in the HEVC development process who can make a reasonable fact-based estimate of when we might see HEVC (or something closely related) show up in a form that can be easily incorporated into Handbrake?
As soon as you write one for us to use.
Please avoid replying if you have no familiarity with HEVC or HEVC development. Thanks.
Please go troll somewhere else.

GregiBoy
Veteran User
Posts: 908
Joined: Sat Feb 12, 2011 9:23 pm

Re: HEVC support? (new standard from MPEG; 40-50% > than H26

Post by GregiBoy » Fri Aug 17, 2012 7:10 am

Wow....@Hippo_Owa

2 posts and 46 hours composing them and we are supposed to be impressed. What a load of codswhallop (B/S)

Either a trool, an academic or a politician!!!

User avatar
Rodeo
HandBrake Team
Posts: 12195
Joined: Tue Mar 03, 2009 8:55 pm

Re: HEVC support? (new standard from MPEG; 40-50% > than H26

Post by Rodeo » Fri Aug 17, 2012 9:36 am

saintdev wrote:Most Blu-rays are still released in MPEG-2.
Well, that one is no longer true (at least when it comes to Hollywood releases).

Smithcraft
Veteran User
Posts: 2691
Joined: Thu Jan 22, 2009 8:04 pm

Re: HEVC support? (new standard from MPEG; 40-50% > than H26

Post by Smithcraft » Fri Aug 17, 2012 10:19 am

I'm not involved with HEVC development, so I guess that makes me an iDiot, but hippo_powa, shouldn't you take this up with the libav folks and not the Handbrake folks?

SC

User avatar
Rodeo
HandBrake Team
Posts: 12195
Joined: Tue Mar 03, 2009 8:55 pm

Re: HEVC support? (new standard from MPEG; 40-50% > than H26

Post by Rodeo » Fri Aug 17, 2012 11:21 am

They're already working on it, too (Libav, that is). But it won't happen overnight. And the real work will probably only begin once the final draft gets released or something.

Smithcraft
Veteran User
Posts: 2691
Joined: Thu Jan 22, 2009 8:04 pm

Re: HEVC support? (new standard from MPEG; 40-50% > than H26

Post by Smithcraft » Fri Aug 17, 2012 12:11 pm

I certainly grasp that, it's just that people keep asking the Handbrake team to do something instead of asking the libav team about it.

SC

dynaflash
Veteran User
Posts: 3820
Joined: Thu Nov 02, 2006 8:19 pm

Re: HEVC support? (new standard from MPEG; 40-50% > than H26

Post by dynaflash » Fri Aug 17, 2012 2:59 pm

hippo_powa wrote:The only useful comment posted so far (in relation to the question being asked) is the comment that 'it'll be ready when someone else does it'. I don't know if that person speaks for the whole handbrake team, but it's at least useful as a starting point.

Now, with all that said. Is there anyone here who has actually been involved in the HEVC development process who can make a reasonable fact-based estimate of when we might see HEVC (or something closely related) show up in a form that can be easily incorporated into Handbrake?


As stated ... unless its implemented in a lib we can use with our license ... and its decided to be worth adding it at *that* time, its mostly mental gymnastics as far as the arguement goes.
hippo_powa wrote:Please avoid replying if you have no familiarity with HEVC or HEVC development. Thanks.
This is not a call that you get to make, unfortunately.

User avatar
JohnAStebbins
HandBrake Team
Posts: 5442
Joined: Sat Feb 09, 2008 7:21 pm

Re: HEVC support? (new standard from MPEG; 40-50% > than H26

Post by JohnAStebbins » Fri Aug 17, 2012 10:08 pm

hippo_powa wrote:Please avoid replying if you have no familiarity with HEVC or HEVC development. Thanks.
If you want to only talk to people knowledgeable in HEVC, you've come to the wrong place. We don't write codecs here. We are essentially integrators. We package together a collection of useful libraries and make using them as fool-proof as we can.

The HEVC spec is schedule to reach final draft in Jan 2013. Then it starts the approval process which can take a while. Now is a good time for someone with the interest in it to start implementation since the spec is probably substantially "final" enough to get a start with and development will take a long time. But it's not something we would work on as a core part of HandBrake. If any of the HandBrake developers had an interest in this kind of work, they would join another project that specializes in it or they would start their own project.

I don't know much about HEVC technical details, but I know plenty about specification processes. HEVC will be a fun science fair project until such time that the spec is final *and* they are successful with their marketing campaign (the latter not being a sure thing). The marketing of HEVC is probably the harder of the 2 tasks. Displacing the billions (probably 100s of billions actually) of dollars in the installed base of equipment that already supports h.264 (and the older mpeg2) is no easy task. DVDs sales *still* beat bluray more than 3 to one. Full adoption of a new standard can take decades. Even a demonstrably superior product has an uphill battle in this environment. It has to be enough better that adopters can foresee a future where trashing all their current equipment results in saving money in the long run. As a concrete example that maybe you can relate to, everyone will have to buy new GPUs that support HEVC because the current ones do not and never will support it. With hard disks getting cheaper *much* faster than GPUs, it may be hard for many to justify upgrading the GPU rather than just buying a bigger HDD.

hippo_powa
Posts: 8
Joined: Thu Aug 16, 2012 1:57 am

Re: HEVC support? (new standard from MPEG; 40-50% > than H26

Post by hippo_powa » Fri Aug 17, 2012 11:42 pm

http://ec2-50-16-227-110.compute-1.amaz ... 0(ppt).pdf

This document may be of interest to anyone curious about HEVC; how it works; when it is happening. There's a technical comparison, benchmarking, and so on. Starts at slide 31.

Notes:

0. Take a look through all the slides, particularly slide 31. The people making these slides have some idea of what they're talking about.
1. They state the July 2012 spec is meant to be 'technically final'; 'technical completion' took place first half 2012.
2. They note that *hardware solutions* will begin to be available during 2013; software sooner. [slide 37]
Last edited by Rodeo on Fri Aug 17, 2012 11:44 pm, edited 1 time in total.
Reason: Fix URL.

hippo_powa
Posts: 8
Joined: Thu Aug 16, 2012 1:57 am

Re: HEVC support? (new standard from MPEG; 40-50% > than H26

Post by hippo_powa » Fri Aug 17, 2012 11:49 pm

"This is not a call that you get to make, unfortunately."

It was not a 'call', it was a polite request to keep the thread on-topic and a polite request for people to contribute on the basis of knowledge rather than ignorance. Those are both good things for any forum thread.

Smithcraft, John: thanks - you're right. I think I have asked in the wrong place. That's OK though. Hopefully some people will have found something interesting in some of the links I posted, and hopefully (since this is the feature request forum) the HB devs will have noticed there are compelling reasons to integrate this into HB as soon as it becomes possible for them to do so.

Thanks for the replies.

hippo_powa
Posts: 8
Joined: Thu Aug 16, 2012 1:57 am

Re: HEVC support? (new standard from MPEG; 40-50% > than H26

Post by hippo_powa » Fri Aug 17, 2012 11:55 pm

Relevant libav mailing list thread.

https://lists.libav.org/pipermail/libav ... 33738.html

hippo_powa
Posts: 8
Joined: Thu Aug 16, 2012 1:57 am

Re: HEVC support? (new standard from MPEG; 40-50% > than H26

Post by hippo_powa » Thu Aug 23, 2012 5:14 pm

First commercial H265 hardware encoder now available and entering real world use.

Anyone still clinging to the idea they won't show up till late 2014? :-)

http://advanced-television.com/index.ph ... c-encoder/

CNET are running an interesting article too, discussing patent issues:

"HEVC, a new weapon in codec wars, to appear in September"
http://news.cnet.com/8301-1023_3-574981 ... september/

Re: decode performance and spec needed to handle the codec:

"Another is that it can divide frames into multiple tiles so multicore processors can spread decoding across parallel subtasks."

This will make adoption rather easier, imho.

User avatar
Rodeo
HandBrake Team
Posts: 12195
Joined: Tue Mar 03, 2009 8:55 pm

Re: HEVC support? (new standard from MPEG; 40-50% > than H26

Post by Rodeo » Thu Aug 23, 2012 5:29 pm

hippo_powa wrote:First commercial H265 hardware encoder now available
OK.
hippo_powa wrote:and entering real world use.
Can you elaborate? Where can you buy it? What software or devices use it to produce HEVC-encoded videos?

Also, where can I get a device to play these HEVC-encoded videos then?

hippo_powa
Posts: 8
Joined: Thu Aug 16, 2012 1:57 am

Re: HEVC support? (new standard from MPEG; 40-50% > than H26

Post by hippo_powa » Fri Aug 24, 2012 6:56 pm

Hello Rodeo

"Can you elaborate? Where can you buy it?"

No, I can't elaborate, I don't work for Ericsson or their customers and I won't be attending that trade show. Did you read the article before posting?

If you're a business user with the right money and need for this particular technology asap, I suggest you read the link I posted and contact their corporate sales team for high end media equipment. They'll almost certainly have the details on expected availability and cost. Why would I have those details? I'm just a hippo. You can ask them to sell you both an encoder and a decoder. I've not dealt with their corporate sales division, so you're on your own there.

Addressing another point of unreasonable skepticism I've seen mentioned recently, elsewhere:

Assertion: "Yeah, but I bet HEVC will only use 50% less data than a really bad implementation of H264"

In reality, Ericsson (who I noted already have this working in *hardware* never mind software), state:

"The compression technology promises to reduce the bandwidth requirements for video delivery by over 50 percent compared to the best H.264/MPEG-4 AVC implementations"
http://www.cio.co.uk/news/3377466/erics ... width-use/

but of course, we can't trust companies...tin foil hats etc... so how about academics. In fact, how about the academic who is generally considered the father of AVC? Well, I suppose he knows very little about video encoding compared to the random troll who mocked academics earlier in this thread, but let's see what he thinks anyway...

""Bin Li, Gary Sullivan and Jizheng Xu published a performance comparison between H.264/AVC and Working Draft 4 of HEVC in November 2011," writes Richardson. Sullivan is considered by many to be the father of H.264, as he sat on both the ITU and MPEG working committees, bringing together MPEG 4 Part 10 and H.264 to what is jointly known as Advanced Video Coding (or AVC).

"On average, HEVC out-performs H.264 by 39% for random access scenarios (e.g. broadcast) and by 44% for low delay scenarios (e.g. video calling)," writes Richardson. "This means that the HEVC codec can achieve the same quality as H.264 with a bitrate saving of around 39-44%.""

http://www.streamingmedia.com/Articles/ ... 84481.aspx

- and that was the performance as of Nov 2011, 3 generations earlier, not Aug 2012.

Mind you, we can hardly trust this mysterious "Richardson" guy to give a fair opinion either, even if he *is* the author of a bunch of textbooks on video codecs... ;-)


Admins, can we move this to another forum? Although it's a feature request, the thread is maybe better placed somewhere relating to general technology discussion, for now.

User avatar
Rodeo
HandBrake Team
Posts: 12195
Joined: Tue Mar 03, 2009 8:55 pm

Re: HEVC support? (new standard from MPEG; 40-50% > than H26

Post by Rodeo » Fri Aug 24, 2012 7:39 pm

hippo_powa wrote:Hello Rodeo

"Can you elaborate? Where can you buy it?"

No, I can't elaborate, I don't work for Ericsson or their customers and I won't be attending that trade show. Did you read the article before posting?

If you're a business user with the right money and need for this particular technology asap, I suggest you read the link I posted and contact their corporate sales team for high end media equipment. They'll almost certainly have the details on expected availability and cost. Why would I have those details? I'm just a hippo. You can ask them to sell you both an encoder and a decoder. I've not dealt with their corporate sales division, so you're on your own there.
My point was just to disagree about your usage of "entering real-world use".

IMO, we can say it "entered real-world use" when HEVC-encoded videos actually reach end-users/consumers (i.e. viewers). Not sure when that'll happen, but not right away.

I won't enter the compression efficiency debate until actual encoder comparisons can be made. Anything else is guesswork (it can be a very good and/or informed guess, but it's still just a guess).

dynaflash
Veteran User
Posts: 3820
Joined: Thu Nov 02, 2006 8:19 pm

Re: HEVC support? (new standard from MPEG; 40-50% > than H26

Post by dynaflash » Tue Aug 28, 2012 7:15 pm

hippo_powa wrote:Admins, can we move this to another forum? Although it's a feature request, the thread is maybe better placed somewhere relating to general technology discussion, for now.
Well, two choices ... Development or Tiki Bar. For now I vote Tiki Bar. :)

At any rate ... tbh someone like Dark_Shikari with x264 and libav is probably more qualified to address this.

He can be found on irc at freenode.net at #libav or better yet #x264 ... though were I you ... I would tread very lightly. Just my .02.

mduell
Veteran User
Posts: 6669
Joined: Sat Apr 21, 2007 8:54 pm

Re: HEVC support? (new standard from MPEG; 40-50% > than H26

Post by mduell » Tue Aug 28, 2012 9:07 pm

Every time any new codec comes up the comparisons all use the same trick: using a lousy existing encoder. In this case your "academic" source used the reference encoder JM 18.0. No one uses JM except for producing compliance samples.

Before HEVC is going to get anywhere it needs to demonstrate superiority over the current state of the art, x264 (with reasonable settings).

Also the current HEVC encoder (HM) is slow. Ungodly slow. Think frames per hour not frames per second. 10+ years to encode an average movie.

hippo_powa
Posts: 8
Joined: Thu Aug 16, 2012 1:57 am

Re: HEVC support? (new standard from MPEG; 40-50% > than H26

Post by hippo_powa » Thu Sep 20, 2012 1:47 am

"Also the current HEVC encoder (HM) is slow. Ungodly slow. Think frames per hour not frames per second. 10+ years to encode an average movie."

mduell, I have no idea why your tag says 'elite bright spark'.

"Vanguard Software Solutions Achieves Full HD at 1080p Software Encoding of HEVC (H.265) Running on Intel x86 Processor

"Vanguard Software Solutions (VSS),founded in 1995 and a leader in H.264/MPEG-4 AVC CODEC technology,announces the world’s first demonstration of a real-time HEVC (H.265) software encoder running at full HD resolution of 1080 progressive scan at 30fps (1080p30). The VSS HEVC encoder is based on the Draft International Standard approved by the Joint Collaborative Team on Video Coding (JCT-VC) in July 2012. To achieve real-time performance at full HD resolution, VSS was able to use a single Intel Xeon processor.

The VSS encoder is optimized for real-time performance on PC platforms, with built-in parallelization for cloud scalability. The encoder demonstrates improvements in quality and bitrate for video distribution and communication markets over current implementations based on H.264/MPEG-4 AVC, achieving lower bitrates than H.264/MPEG-4 AVC while maintaining similar quality levels."

Read more here: http://www.heraldonline.com/2012/09/09/ ... rylink=cpy

but, you know, don't let the facts get in the way of a good trolling.

User avatar
JohnAStebbins
HandBrake Team
Posts: 5442
Joined: Sat Feb 09, 2008 7:21 pm

Re: HEVC support? (new standard from MPEG; 40-50% > than H26

Post by JohnAStebbins » Thu Sep 20, 2012 8:32 am

I expect there will be plenty more such announcements. But I can find absolutely nothing regarding 3rd party reviews or comparisons between their encoder and any other encoder. So all we know at this point is the company that produce the encoder thinks it's the hottest thing since sliced bread. No surprises there. It could be implementing a minimal subset of the standard to achieve the necessary real-time speed for all we know. It could well be both slower than x264 and have lower quality than x264 for all we know. There just isn't sufficient information yet.

I attended IBC and talked to several broadcasters and decoder box manufacturers about their roll-out plans for HEVC. None are forecasting earlier than late 2013. Since these large projects are always late, I would expect to see small test markets getting HEVC broadcasts in early to mid 2014. This kind of upgrade is usually rolled out in low volume high-end systems first and slowly over the coarse of about a decade percolates out to the reset of their customers.

I could see this taking hold in consumer gadgets like smartphones and protable media players a bit faster. But there is a coordination effort involved here as well. I'm not as well tied in to this market, so I don't know the product development cycle as well. But I don't recall seeing any announcements for smartphone chips with hevc support yet.

These things always take longer than you would expect to really start to penetrate the market.

Post Reply