A Proposal to Accelerate HEVC Adoption

[x265_Project]

Today I'm publishing a proposal to accelerate HEVC adoption. I look forward to a productive conversation.
http://x265.org/proposal-accelerate-hevc-adoption/

Tom Vaughan
VP and GM, Video
MulticoreWare

[JohnLai]

The problem with hardware HEVC encoder, most of the time, it produces similar bitrate and quality to H264 hardware encoder. Probably due to a lot of missing HEVC features (such as SAO, B-Frame or I-Frame, limited reference frames support, lack of larger or smaller Partition Unit Size). It is all pure speed with 'acceptable' quality and file size. Not to mention one of the world largest player --> Youtube doesn't actually accept HEVC as input stream yet. Same with Twitch.tv

Google for one......they would be hesitate to accept hevc stream. (Patent problem) Beside, google spent quite a lot on promoting and readying its infrastructure for VP9.

Software encoder such as x265 supports almost all of HEVC features and produce highest quality and smallest file size (best compression efficiency). Unfortunately, with all features being enabled, the encoding speed greatly suffers (1-3fps for 1080p content).
Assuming if majority of gamers + streamers with intel 3570k, 4670k or 6600k use x265 for encoding 1080p30 content (One doesn't expect them to use extremely expensive multisocket Intel Xeons CPUs.....), they are forced to use Ultrafast preset (with majority of hevc defining features missing) for reliable real time encode.
Not to mention even with ultrafast preset, the high load on cpu will ruin users gaming experience.

[x265_Project]

The problem with hardware HEVC encoder, most of the time, it produces similar bitrate and quality to H264 hardware encoder. Probably due to a lot of missing HEVC features (such as SAO, B-Frame or I-Frame, limited reference frames support, lack of larger or smaller Partition Unit Size). It is all pure speed with 'acceptable' quality and file size. Not to mention one of the world largest player --> Youtube doesn't actually accept HEVC as input stream yet. Same with Twitch.tv

Google for one......they would be hesitate to accept hevc stream. (Patent problem) Beside, google spent quite a lot on promoting and readying its infrastructure for VP9.

Software encoder such as x265 supports almost all of HEVC features and produce highest quality and smallest file size (best compression efficiency). Unfortunately, with all features being enabled, the encoding speed greatly suffers (1-3fps for 1080p content).
Assuming if majority of gamers + streamers with intel 3570k, 4670k or 6600k use x265 for encoding 1080p30 content (One doesn't expect them to use extremely expensive multisocket Intel Xeons CPUs.....), they are forced to use Ultrafast preset (with majority of hevc defining features missing) for reliable real time encode.
Not to mention even with ultrafast preset, the high load on cpu will ruin users gaming experience.
My blog post is focused on the patent licensing situation. HEVC patent holders should understand that if HEVC is adopted widely, consumers will look for native hardware decoding in their new devices at a minimum. Native hardware encoding is also very valuable for real-time, low latency encoding, such as video chat/conferencing, screen sharing (WebEx, GotoMeeting), remote graphics (VMWare and other virtualized machine environments), and other mobile applications (recording video on your smartphone).

[littlepox]

I don't think a lot of commercial entities are crying for HEVC unless 4K becomes widely available. They have been comfortable with AVC @ 1080p for years. Recently a Chinese online video website is happy with its improvements to "cut down the bandwidth to half"; guess how? They steal x264 and replace whatever poor encoder they used previously. And their audience are happy as well since they do see the "improvements"

[mariush]

I think the best for adoption would be some company or group of companies to just pay the money and release a plugin like the OpenH264 Video Codec that Cisco makes. They could afford to pay the millions of dollars Not sure if current licensing schemes allow something like that.

I bet that if browsers like Firefox would stream hevc in html5 player natively, porn sites would almost instantly accept it (better quality, less bitrate ... what's not to like)

[JohnLai]

I don't think a lot of commercial entities are crying for HEVC unless 4K becomes widely available. They have been comfortable with AVC @ 1080p for years. Recently a Chinese online video website is happy with its improvements to "cut down the bandwidth to half"; guess how? They steal x264 and replace whatever poor encoder they used previously. And their audience are happy as well since they do see the "improvements"

Like it or not......the truth is....piracy/theft will be a main factor for major HEVC adoption.

Similar in a way when MP3 was first introduced in era of dial-up/dsl.
AAC has hard time replacing MP3 as de-facto audio format until recently.
Patent protection basically did nothing to stiffle adoption of MP3 in the past.

[x265_Project]

I don't think a lot of commercial entities are crying for HEVC unless 4K becomes widely available. They have been comfortable with AVC @ 1080p for years. Recently a Chinese online video website is happy with its improvements to "cut down the bandwidth to half"; guess how? They steal x264 and replace whatever poor encoder they used previously. And their audience are happy as well since they do see the "improvements"

If a website is using x264 in their own servers, they can use it for free under the GPL v2 license. That's not stealing. This is one reason that x264 became so popular - it's used by YouTube, Netflix, Hulu, Facebook, and most of the leading web services around the world.

[x265_Project]

Like it or not......the truth is....piracy/theft will be a main factor for major HEVC adoption.

Similar in a way when MP3 was first introduced in era of dial-up/dsl.
AAC has hard time replacing MP3 as de-facto audio format until recently.
Patent protection basically did nothing to stiffle adoption of MP3 in the past.

I disagree. Legitimate companies don't "steal" codecs. Unless major web services and app developers are comfortable adopting HEVC in their apps and services, it won't succeed broadly. No amount of ripping and torrenting is going to help HEVC succeed. What we need is for every consumer in the world to be comfortable saving and sharing video as HEVC, because we can count on everyone else being able to handle that format, just as we do with any MP4 file today. That won't happen unless HEVC support becomes native on the leading consumer device platforms... Android, iOS, Windows, MacOS. Device OEMs don't want to sponsor this cost if there is no demand - so my proposal is designed to stimulate this demand. When app developers and web services are free to use HEVC, it will make sense for device OEMs to support it, and they'll be happy to pay the royalties involved, as it will be a strong selling point for their devices.

[littlepox]

If a website is using x264 in their own servers, they can use it for free under the GPL v2 license. That's not stealing. This is one reason that x264 became so popular - it's used by YouTube, Netflix, Hulu, Facebook, and most of the leading web services around the world.

wow... did not know that; I thought one must pay for x264 if he is using for commercial purposes. Good to know they are actually doing all right.
I'd assume the same applies to x265 since x265 is offered under the same license.

we can count on everyone else being able to handle that format, just as we do with any MP4 file today.

MP4 files may actually be encoded in HEVC format nowadays; if we can handle "any" of them, we can handle all profiles of HEVC already:p


When app developers and web services are free to use HEVC, it will make sense for device OEMs to support it, and they'll be happy to pay the royalties involved

But the converse is also true: they will be skeptical for HEVC unless OEMs support is ready. This is one reason why 10bit-AVC has never been successful only except for community use, and same for vorbis, which is completely free.


No amount of ripping and torrenting is going to help HEVC succeed.

Ironically, those pirates and their followers have always been the pioneers for new codec. It has been years that we see x265 being deployed for those sweet-pot encodes, and nowadays it is just becoming more popular. A lot of useful feedback and critics you've collected have been from them. But I do agree their impact is low, otherwise High 10 Profile AVC should have been dominating.

[hajj_3]

Not going to happen. Even if all patent holders agree to all these terms then it still won't be widely used on the internet due to AV1 being just 4-7 months away from being ratified and 18-24 months away from being used by all the big online video websites. Companies will switch to AV1 for devices with hardware decoders and for pc users that manually enable it, they will used vp9 as a fallback codec for those with a vp9 hardware decoder, maybe even for those with only a software decoder. For those with neither an AV1 or VP9 hardware decoder they will use h264 or will just use software vp9 decoder. Netflix wrote on their blog yesterday that vp9 is great, it looks like they will be adopting it for mobile devices with a hardware decoder. It sounds like the AOM will likely release a new codec every 3yrs or so which will mean that MPEG will likely never have a codec used on the internet in the future, no online companies will adopt h266 or h267.

Intel's kaby lake chips have vp9 10bit hardware decoders and even 8bit hardware encoders so there is even less reason not to adopt vp9.

[FancyMouse]

wow... did not know that; I thought one must pay for x264 if he is using for commercial purposes. Good to know they are actually doing all right.
I'd assume the same applies to x265 since x265 is offered under the same license.
That's exactly why there's another AGPL license.

[smegolas]

But I do agree their impact is low, otherwise High 10 Profile AVC should have been dominating.

Bad example because H.264 Hi10P is used by maybe 1% of pirates.

[x265_Project]

Netflix wrote on their blog yesterday that vp9 is great, it looks like they will be adopting it for mobile devices with a hardware decoder.
Netflix did not write on their blog that "VP9 is great". They gave a high level summary of a study that they presented in more detail today (https://youtu.be/wi1BefrfTos?t=1h25s). Here's a summary comparing codec efficiency using Netflix's advanced video quality metric...http://x265.org.s3.amazonaws.com/img/Netflix_Codec_Comparison_Results_50.png

I am in the industry, and I work with many industry leaders. I know what they thinking, because they tell me what they are thinking. I can't speak for anyone but MulticoreWare and our x265 project. But I think the way to think of how a movie streaming service views codecs is that when a customer wants to watch a movie on a device, the movie streaming service wants to deliver the highest quality experience possible on that device, regardless of network conditions. If that device supports H.264 and H.265, but not VP9, what codec does a competitive movie streaming service want to use? If bandwidth isn't an issue, you can just use H.264 and more bandwidth. But if bandwidth is not cheap and plentiful, H.265 will deliver a better quality experience.

Sure, licensing costs are an issue, and that is why I'm pushing to make licensing costs more reasonable and more certain. Codecs take years to develop, optimize, and deploy. AV1 will almost certainly be a great option 3 or 4 years from now. At the moment, it isn't a final standard. Video services need to compete today, tomorrow and next year. They don't want to lose to a competitor while they're waiting for something better. Of course, in the long term, they don't want to be hostage to licensing terms that are detrimental to their business, and so we see large movie streaming services working to promote and develop better long-term solutions.

[aymanalz]

I bet that if browsers like Firefox would stream hevc in html5 player natively, porn sites would almost instantly accept it (better quality, less bitrate ... what's not to like)

Why just porn sites? Wouldn't the same reasons apply to any other streaming websites, including youtube and hulu and netflix?

[Atak_Snajpera]

Netflix did not write on their blog that "VP9 is great". They gave a high level summary of a study that they presented in more detail today (https://youtu.be/wi1BefrfTos?t=1h25s). Here's a summary comparing codec efficiency using Netflix's advanced video quality metric...http://x265.org.s3.amazonaws.com/img/Netflix_Codec_Comparison_Results_50.png

I am in the industry, and I work with many industry leaders. I know what they thinking, because they tell me what they are thinking. I can't speak for anyone but MulticoreWare and our x265 project. But I think the way to think of how a movie streaming service views codecs is that when a customer wants to watch a movie on a device, the movie streaming service wants to deliver the highest quality experience possible on that device, regardless of network conditions. If that device supports H.264 and H.265, but not VP9, what codec does a competitive movie streaming service want to use? If bandwidth isn't an issue, you can just use H.264 and more bandwidth. But if bandwidth is not cheap and plentiful, H.265 will deliver a better quality experience.

Sure, licensing costs are an issue, and that is why I'm pushing to make licensing costs more reasonable and more certain. Codecs take years to develop, optimize, and deploy. AV1 will almost certainly be a great option 3 or 4 years from now. At the moment, it isn't a final standard. Video services need to compete today, tomorrow and next year. They don't want to lose to a competitor while they're waiting for something better. Of course, in the long term, they don't want to be hostage to licensing terms that are detrimental to their business, and so we see large movie streaming services working to promote and develop better long-term solutions.

I wish my eyes could verify those numbers ;) My brain/eyes tell me that bitrate saving is more around 25% than 50%. Unfortunately the amount of time needed for that extra 25% is just ridiculous on 8C/16T CPU! x265 preset slower was around 4 times slower than x264 veryslow. Also I'm not big fan of that weird zigzag/oil paint effect on x265 samples. In my opinion those claims that you can save half of bandwidth without quality loss is just another marketing lie. I'm not surprised that industry hasn't jumped so enthusiastically on this h.265 train like they did with h.264.

x264 10bit veryslow 2000kbps (https://mega.nz/#!wMdGTA5S!H1phCxTv9eBe4WZ74_8G4gM3LI9ujVlFSpx2il2imfQ) 2-pass
http://i.cubeupload.com/yslsUa.png
http://i.cubeupload.com/XgBwTw.png
http://i.cubeupload.com/sqil8P.png
http://i.cubeupload.com/bHhDp8.png

x265 10bit slower 1000kbps (https://mega.nz/#!gI1mDbRB!FsofBaoywp0gwuJYEz2EZ2xCyOhO4NJF2Fusq-m-qCE) 2-pass
http://i.cubeupload.com/FQkglt.png
http://i.cubeupload.com/zZds9T.png
http://i.cubeupload.com/szHynJ.png
http://i.cubeupload.com/SLl0PZ.png

[Khanattila]

Netflix did not write on their blog that "VP9 is great". They gave a high level summary of a study that they presented in more detail today (https://youtu.be/wi1BefrfTos?t=1h25s). Here's a summary comparing codec efficiency using Netflix's advanced video quality metric...http://x265.org.s3.amazonaws.com/img/Netflix_Codec_Comparison_Results_50.png

I am in the industry, and I work with many industry leaders. I know what they thinking, because they tell me what they are thinking. I can't speak for anyone but MulticoreWare and our x265 project. But I think the way to think of how a movie streaming service views codecs is that when a customer wants to watch a movie on a device, the movie streaming service wants to deliver the highest quality experience possible on that device, regardless of network conditions. If that device supports H.264 and H.265, but not VP9, what codec does a competitive movie streaming service want to use? If bandwidth isn't an issue, you can just use H.264 and more bandwidth. But if bandwidth is not cheap and plentiful, H.265 will deliver a better quality experience.

Sure, licensing costs are an issue, and that is why I'm pushing to make licensing costs more reasonable and more certain. Codecs take years to develop, optimize, and deploy. AV1 will almost certainly be a great option 3 or 4 years from now. At the moment, it isn't a final standard. Video services need to compete today, tomorrow and next year. They don't want to lose to a competitor while they're waiting for something better. Of course, in the long term, they don't want to be hostage to licensing terms that are detrimental to their business, and so we see large movie streaming services working to promote and develop better long-term solutions.

PSNR-tuned configuration, the results are rubbish.
Visual Quality configuration, more interesting.

EDIT. Why not try even larger resolutions?
In addition, the MS-SSIM data can not be simply ignored.

[x265_Project]

The topic of this thread is HEVC patent licensing terms, and the need for HEVC patent holders to find ways to accelerate HEVC adoption.

[MeteorRain]

Bad example because H.264 Hi10P is used by maybe 1% of pirates.

That's the result. Since no device support hardware decoding Hi10P, rippers have to abandon it.

[raffriff42]

I see your proposal was republished at streamingmedia.com (http://www.streamingmedia.com/Articles/Editorial/Featured-Articles/Its-Time-to-Move-Forward-with-HEVC-113278.aspx), without comment unfortunately. It's a very worth proposal (it seems to me, as an outsider) which can perhaps be summed up as "let's be reasonable with the licensing fees, guys." Which is great, and I hope the HEVC patent holders listen to you.

Your article seems to be resigned to the eventual adoption of AV1. That's the impression I'm getting as well. It's hard to fight AMD, ARM, Amazon, Cisco, Google, Intel, Microsoft, Mozilla, NVIDIA and Netflix when they all get together like this. I think perhaps the HEVC patent holders are thinking the same, and looking forward to their next revenue model: lawsuits! Also, FUD! (Untangling the possible patent infringement exposure in AV1 could obviously take many years; HEVC can collect billions in fees in the meantime, even if the lawsuits eventually fail)

[smegolas]

That's the result. Since no device support hardware decoding Hi10P, rippers have to abandon it.
99% of encoders never even considered to use it, since Hi10P is not a consumer profile and was never intended for consumer use.

It was only the asian cartoon fans who used it. And this is such a niche market (mostly children I guess?) that it cannot influence hardware trends.

So it is bad example to say that piracy cannot influence hardware adoption based on some niche cartoon fans using something that was never intended to be adopted. Whereas HEVC Main and Main 10 are intended for consumers since their very inception.

[x265_Project]

I see your proposal was republished at streamingmedia.com (http://www.streamingmedia.com/Articles/Editorial/Featured-Articles/Its-Time-to-Move-Forward-with-HEVC-113278.aspx), without comment unfortunately. It's a very worth proposal (it seems to me, as an outsider) which can perhaps be summed up as "let's be reasonable with the licensing fees, guys." Which is great, and I hope the HEVC patent holders listen to you.

Your article seems to be resigned to the eventual adoption of AV1. That's the impression I'm getting as well. It's hard to fight AMD, ARM, Amazon, Cisco, Google, Intel, Microsoft, Mozilla, NVIDIA and Netflix when they all get together like this. I think perhaps the HEVC patent holders are thinking the same, and looking forward to their next revenue model: lawsuits! Also, FUD! (Untangling the possible patent infringement exposure in AV1 could obviously take many years; HEVC can collect billions in fees in the meantime, even if the lawsuits eventually fail)
Thanks! The HEVC patent holders are listening. Some more than others.

Yes, I think the Alliance for Open Media will be successful. The AOM represents some of the largest users of video codecs coming together to say "we won't pay exorbitant patent royalties". In the long run, it's cheaper for these companies to develop the best possible codec using all available techniques that are either based on patents that have expired, or based on techniques that companies and organizations are willing to contribute royalty free. Video codecs are mission-critical core technology for these companies (which include many of the biggest tech companies in the world). They can't afford to be held hostage by a collection of smaller companies.

This is why it's critical for HEVC patent holders to wake up and smell the coffee burning. There's no time to waste.

As for HEVC patent holders suing AV1 implementers... it's a possibility, but I'm not sure any sane company wants to take on Google, Microsoft, Intel, Amazon, Netflix and many others at the same time. These companies have huge patent portfolios that they can use both defensively and offensively, and plenty of lawyers working for them. In addition, they could attack the validity of HEVC patents, invalidating patents that are obvious, or based on prior art. So while HEVC patent holders would definitely like to create some fear, uncertainty and doubt about the royalty free status of AV1, it's hard to say (without seeing all the specific patents and any supposedly infringing techniques in AV1) what their chances are. You can be sure the AOM is working hard to avoid this kind of problem, and that they know what they're doing when it comes to vetting the origin of every encoding technique contributed.

[hajj_3]

I think perhaps the HEVC patent holders are thinking the same, and looking forward to their next revenue model: lawsuits! Also, FUD! (Untangling the possible patent infringement exposure in AV1 could obviously take many years; HEVC can collect billions in fees in the meantime, even if the lawsuits eventually fail)

It say in the video posted last week at the conference that hevc said that they are looking at extending the hevc spec to add support for 256x256 instead of 64x64 for one of the things that the codec does (can't remember what). They said that this would improve compression by either 17% or 27% or something. This would be at the expense of increased computation of course. Clearly the hevc companies are scared of AV1 otherwise they wouldn't be adding to the spec which would break compatibility with existing hardware decoders. I expect that they will probably try and ratify a completely new codec in 3yrs time that would be better than AV1 in the hope that companies will adopt their patented codec. I believe this would fail as the AOM would just make an AV2 codec to compete with that. The days of patented codecs on the web are numbered :) I'm sure patented codecs will still be used in broadcast though.

[Khanattila]

99% of encoders never even considered to use it, since Hi10P is not a consumer profile and was never intended for consumer use.

It was only the asian cartoon fans who used it. And this is such a niche market (mostly children I guess?) that it cannot influence hardware trends.

So it is bad example to say that piracy cannot influence hardware adoption based on some niche cartoon fans using something that was never intended to be adopted. Whereas HEVC Main and Main 10 are intended for consumers since their very inception.

99% of encoders would not know how to tie his own shoes.

[smegolas]

99% of encoders would not know how to tie his own shoes.

And how is that relevant to my point?

Someone said that piracy cannot influence adoption of a codec and used anime Hi10P as proof.

It's a poor example because this is a very niche/nerd market, most people do not watch this stuff.

If a large proportion of the online encodes for popular mainstream content (live action movies/shows etc) were suddenly HEVC then it would surely increase demand and ownership of HEVC compatible hardware.

Installed hardware is one factor that legit streaming companies will consider when choosing their roll-out strategy.

[MoSal]

Someone said that piracy cannot influence adoption of a codec and used anime Hi10P as proof.

It's a poor example because this is a very niche/nerd market, most people do not watch this stuff.


It's worth noting that ,other than people who read specs, almost no one even heard about Hi10P before x264 started to implement it. And that was way after H264/x264 started to become ubiquitous. Around the same time, FFmpeg added a software decoder. Hi10P wasn't, and wouldn't be, even playable in most devices without that decoder.

The story of Hi10P adoption could have been dramatically different if it was supported earlier.

If a large proportion of the online encodes for popular mainstream content (live action movies/shows etc) were suddenly HEVC then it would surely increase demand and ownership of HEVC compatible hardware.

HEVC is actually popular among people who live with data caps. "Over-compressed encodes with good enough quality" is one of the few valid use-cases of x265. 4K and 2D anime are probably valid use-cases too. But I don't have first-hand experience to attest to that.

HEVC encoders are still not good enough at producing high quality 1080p/720p encodes. There are cases where x265 produces slightly better results (at the same bitrate) than x264. But the occasional (yet small) efficiency wins, and the inconsistency makes it unjustifiable to use it over the more compatible H264.

[benwaggoner]

Clearly the hevc companies are scared of AV1 otherwise they wouldn't be adding to the spec which would break compatibility with existing hardware decoders.
In my experience MPEG mainly worries about competing with their last generation codec. MPEG itself doesn't deal with all the patent licensing stuff (for better or worse). And they've always done revs of codecs to add new functionality. MPEG-4 and H.264 saw several rounds of extensions like we're seeing with HEVC, generally to extend beyond basic consumer delivery to professional scenarios and scalable transmission.

[smegolas]

It's worth noting that ,other than people who read specs, almost no one even heard about Hi10P before x264 started to implement it. And that was way after H264/x264 started to become ubiquitous. Around the same time, FFmpeg added a software decoder. Hi10P wasn't, and wouldn't be, even playable in most devices without that decoder.

The story of Hi10P adoption could have been dramatically different if it was supported earlier.



Hi10P adoption would not have been dramatically different because it was never intended to be a consumer profile. Therefore a decoding chipset was never on any hardware vendors roadmap.

For chipset design, it takes several years to go from planning stage to market. Do you really think major vendors would have made the costly decision to incorporate Hi10P into their chipsets? All because a small number of kids were using it for their cartoons? It was never going to happen.

Especially not because HEVC was in the planning stages since 2007. Everybody knew it was coming. So that was even more of a reason not to incorporate Hi10P decoding in chipsets because everybody knew this fad would be very short lived.

[Khanattila]

And how is that relevant to my point?

Someone said that piracy cannot influence adoption of a codec and used anime Hi10P as proof.

It's a poor example because this is a very niche/nerd market, most people do not watch this stuff.

If a large proportion of the online encodes for popular mainstream content (live action movies/shows etc) were suddenly HEVC then it would surely increase demand and ownership of HEVC compatible hardware.

Installed hardware is one factor that legit streaming companies will consider when choosing their roll-out strategy.

I mean that many people do not know what 8-bit and 10-bit are, or 4:2:2 and all other things. They take a script that randomly by someone else without getting a lot of questions.

EDIT. Since there is Twitch, I see a slight majority interest for some advanced option, but most of the time, people copy settings from a web page.

[Motenai Yoda]

@MoSal I can say x265 actually produces comparable "high quality" files at about 70%-80% lower bitrate than x264, and on mid quality even 50%-60%

[MoSal]

@MoSal I can say x265 actually produces comparable "high quality" files at about 70%-80% lower bitrate than x264.

Sounds plausible, If you're setting >7m away from a 21" screen ;)

Joking aside. My eyes tend to be very sensitive to blur. Even x264's default tunings are not good enough for me (unless low CRF values are used of course). The blurriness is bad enough in many videos that I have to enable a relatively weak sharpening filter in my player to view them without cursing.

`--deblock -2:-2` in x264 produces great results for my eyes. It's usually enabled via `--tune grain`. Some rippers take it too far and use `--deblock -3:-3`. But that's a little bit too much for videos.

The last time I tested x265, no preset/tuning combination produced good enough results for my eyes (within the bitrate range usually used for high-quality online content), to justify using it over a tuned x264.

I don't think my eyes are unique in any way. As many users here in this forum report similar experiences.

-----

Back on-topic. One use-case that I didn't see many people take advantage of is high-delay live streams (HLS, DASH, ... etc). The range of quality offered in those streams for demanding content (e.g. sports events) usually ranges from barely acceptable to almost good.

I stumbled upon an (HLS-like) live stream once that was 2250kbps/30fps/720p. I remember being surprised by the good quality offered. I checked my player to see if the bitrate reported was correct. And that's when I discovered that HEVC was actually used for the video. That was the first and last HEVC-enabled live stream I ever stumbled upon.

[pandy]

My 2 cents as usual:

Without price affordable large 4k displays i see no significant gain from introducing H.265 when compared to less computationally intense codecs.

H.265 adoption rate is highly linked with receivers and displays capable to decode and display 4k content.
85 inch 4k TV's are still too expensive for average casual customer.

[benwaggoner]

My 2 cents as usual:

Without price affordable large 4k displays i see no significant gain from introducing H.265 when compared to less computationally intense codecs.

H.265 adoption rate is highly linked with receivers and displays capable to decode and display 4k content.
85 inch 4k TV's are still too expensive for average casual customer.
Prices per inch for large displays have been dropping dramatically year-on-year. The introduction of HDR at the high end has kept high-end TV prices from doing that, but that's because they are getting 2-3x brighter and awesome in lots of other ways. A basic 65" SmartTV with support for HEVC Main 10 2160p60 playback is way less than it was two years ago.

Also, HEVC offers big efficiency improvements for mobile, and lots of SoCs include HEVC decode now. That's a whole separate and important market for HEVC. And one where >bits is >cost, so there's a financial incentive for better compression, and customers don't really expect "perfect." But if you only have 300 Kbps to spend on SD video, HEVC can way outperform H.264.

[pandy]

Prices per inch for large displays have been dropping dramatically year-on-year. The introduction of HDR at the high end has kept high-end TV prices from doing that, but that's because they are getting 2-3x brighter and awesome in lots of other ways. A basic 65" SmartTV with support for HEVC Main 10 2160p60 playback is way less than it was two years ago.

True but i think about 85 and more inch - prices are quite high and i don't see significant benefits to have 4k on 65 inch...


Also, HEVC offers big efficiency improvements for mobile, and lots of SoCs include HEVC decode now. That's a whole separate and important market for HEVC. And one where >bits is >cost, so there's a financial incentive for better compression, and customers don't really expect "perfect." But if you only have 300 Kbps to spend on SD video, HEVC can way outperform H.264.

Mobile is out of my scope as 4k H.265 (OK 1k H.265) sound hilarious on 5.5 max inch screen...
Add to this VFR and mobile is for me like MacDonald for food - consuming quantitatively content without quality...

Btw - i really miss argument about HDR on mobile...
https://www.youtube.com/watch?v=S_8OPi9iqBU Home cinema for all - we need 4k and HDR on mobile .

[benwaggoner]

These is veering off topic, and I justify responding only in that this is explanatory for what the markets for HEVC are and their scope.

True but i think about 85 and more inch - prices are quite high and i don't see significant benefits to have 4k on 65 inch...

Depends on how far away you sit. But HDR is valuable from any distance, as is much better quality at low bitrates (WiFi when the microwave is making popcorn for the movie...). Also, 10-bit HEVC is more common than 10-bit H.264.

Mobile is out of my scope as 4k H.265 (OK 1k H.265) sound hilarious on 5.5 max inch screen...
Add to this VFR and mobile is for me like MacDonald for food - consuming quantitatively content without quality...
Mobile includes large tablets and even laptops. And for every screen size, being able to get a better experience at available bandwidth (or lower bandwidth costs) is valuable.

Btw - i really miss argument about HDR on mobile...
https://www.youtube.com/watch?v=S_8OPi9iqBU Home cinema for all - we need 4k and HDR on mobile .
Check out the Amazon Video HDR demos on a Note 7. HDR provides a massive experience upgrade even on that size of a screen.

[pandy]

These is veering off topic, and I justify responding only in that this is explanatory for what the markets for HEVC are and their scope.

Thx Ben - as i respect you then i need to clarify that it was not personal and my point was only to show that H.265 is designed to be efficient only on 4k progressive content, with lower resolutions it will be not able to deliver promised half bitrate reduction with same quality, 10 bit as standard is most important thing brought by H.265 to customers.
I enjoy HDR but to use fully HDR you need to have proper environment (light!) and as such mobile will be unable to fully use HDR content unless such conditions are fulfilled.

Btw if you experiencing microwave oven impact on WLAN (WiFi) performance then i would recommend to change microwave - seriously there is probably something wrong with oven and energy leaking from microwave is a serious health risk.

[benwaggoner]

Thx Ben - as i respect you then i need to clarify that it was not personal and my point was only to show that H.265 is designed to be efficient only on 4k progressive content, with lower resolutions it will be not able to deliver promised half bitrate reduction with same quality, 10 bit as standard is most important thing brought by H.265 to customers.
10-bit is great, although unfortunately not available for lots of devices where low bitrate matters.

Honestly, we may be getting close to half bitrate v. H.264 for very low bitrates. HEVC degrades with high QP a lot more gracefully than H.264 (which in turn degraded much more gracefully than MPEG-2). At the bitrates required for fallback over poor cellular connections, HEVC can do some amazing things.


I enjoy HDR but to use fully HDR you need to have proper environment (light!) and as such mobile will be unable to fully use HDR content unless such conditions are fulfilled.
Again, try the Note 7. But one of the great things about a phone v. TV is that you can carry it where you want to and adjust the angle perfectly. Lots of people watch video on mobile in dark environments with no glare; probably better on average than for TVs.



Btw if you experiencing microwave oven impact on WLAN (WiFi) performance then i would recommend to change microwave - seriously there is probably something wrong with oven and energy leaking from microwave is a serious health risk.

*I* don't have that problem - all my home streaming devices are connected via Cat6 GigE :). But it is a more common customer problem than you might think. Microwaves and older WiFi are both 2.4 GHz, and people like to microwave popcorn while the credits roll. If the microwave is between the base station and the playback device, bandwidth can drop like a stone.




Sent from my iPad using Tapatalk

[Asmodian]

I do find it odd that the patent holders are being so unreasonable that HEVC is not seeing rapid general adoption. Encouraging companies like Netflix to create their own open codec could destroy the entire codec licensing business for everyone.

I wonder if this issue with HEVC will be possible to fix at this point? Did this threat of content based licensing costs, that started the open codec alliance, permanently push Netflix and YouTube into an open codec only world view? Will these big players continue to use licensed codecs in the future, even if they have the same licence fees as AVC?

Btw if you experiencing microwave oven impact on WLAN (WiFi) performance then i would recommend to change microwave - seriously there is probably something wrong with oven and energy leaking from microwave is a serious health risk.

Sorry for more microwave talk but this is a pet peeve of mine. :o

Panasonic "Inverter" microwaves hurt WiFi much more than "normal" microwaves. All microwaves blast the 2.4 GHz range with large amounts of random radiation but normal ones do it with a 60 Hz on-off cycle so WiFi manages to get some data through in the gaps. However, Inverter microwaves use much higher frequency pulses (1000Hz) which does not leave enough room between pulses for WiFi to get much data through. This does make them better at heating food though. :p

[pandy]

10-bit is great, although unfortunately not available for lots of devices where low bitrate matters.

And it will not change quickly at least in Europe as for example Germany decided to implement 8 bit H.265 for terrestrial broadcast and i'm afraid that they may have lot of followers in Europe... (but with 4k this may be less important issue as with help of dithering banding may be reduced) - side to this it is quite strange as we all knows that higher bit depth usually lead to bitrate reduction.


Honestly, we may be getting close to half bitrate v. H.264 for very low bitrates. HEVC degrades with high QP a lot more gracefully than H.264 (which in turn degraded much more gracefully than MPEG-2). At the bitrates required for fallback over poor cellular connections, HEVC can do some amazing things.

Maybe but with 4k content where large CTU may be beneficial for small resolution i doubt it can even match this kind of level for bitrate reduction... it was never case for H.264 to be so beneficial when compared to previous source coding methods and from my perspective H.265 is not so different form H.264 - mostly brute force scaled up.


Again, try the Note 7. But one of the great things about a phone v. TV is that you can carry it where you want to and adjust the angle perfectly. Lots of people watch video on mobile in dark environments with no glare; probably better on average than for TVs.

Have no access to Amazon demos... can't try it.
And dark environment are bad for eyes...



*I* don't have that problem - all my home streaming devices are connected via Cat6 GigE :). But it is a more common customer problem than you might think. Microwaves and older WiFi are both 2.4 GHz, and people like to microwave popcorn while the credits roll. If the microwave is between the base station and the playback device, bandwidth can drop like a stone.

As electronics engineer involved in WiFi (also performance characterization from video streaming perspective) i'm fully aware of all this.

People problem with 2.4GHz is many different transmissions (2.4GHz is a ISM band) and if adding to this usually low quality for WiFi itself (series of limitations due physical nature of WiFi but also because products are designed not by engineers - currenlty accountants are more important) nobody can expect same quality of link as in case of cable - this should be clear message for customers - marketing reality is completely opposite.

To conclude - thumbs up for H.265 but i see no way to accelerate H.265 adoption - so many things must happen before.


Sorry for more microwave talk but this is a pet peeve of mine. :o

Panasonic "Inverter" microwaves hurt WiFi much more than "normal" microwaves. All microwaves blast the 2.4 GHz range with large amounts of random radiation but normal ones do it with a 60 Hz on-off cycle so WiFi manages to get some data through in the gaps. However, Inverter microwaves use much higher frequency pulses (1000Hz) which does not leave enough room between pulses for WiFi to get much data through. This does make them better at heating food though. :p

IMHO there is something wrong with device - there should be no energy leak from oven...

[JohnLai]

Speaking about patent, how about current competing codec such as China's AVS+(AVS 2.0) (China is a huge market)? This codec is mandatory for China market as its government has outlined every broadcaster must support AVS+ for terrestrial and satellite broadcast.
Some may accused AVS+ of copying other organization patent, but China insists it has developed its own patented codec to reduce reliance on foreign codec ---> due to high licensing fee and patent litigation.

I don't think China broadcaster gonna use HEVC at this point.

[iwod]

Is this too little too late? I asked for EXACTLY the same thing when MPEG-LA announced it was going to be a 25M cap for HEVC. That was many years ago, 5 times the rate of AVC. But i guess that was fine as time goes on, the economy has changed since AVC.
As mobile phones were taking over the world, we end up with a few big players owning 80-90% of market. I am not sure if this was the case, where a few company are now shipping the majority, therefore hitting the cap much easier, and the total revenue for HEVC to MPEG-LA dropped, some company decided it wasn't fair or were being greedy, decide to start another patent pool. HEVC Advance. ( Actually a few large companies including Microsoft* were left out of the MPEG-LA pool from the start )

Their first offer was a ridiculous fees with no cap, I posted a few response offering pretty much the same as the blog post. The HEVC Advances 2nd offering, which is still a ridiculous 40M cap. I posted the same thing again, I think it was on Streamingmedia. Now that it is none of my business I would really want it to die. Only regret is I dont get to see Bellard's BPG takes off, ( Which is based on HEVC )

At one stage i thought it didn't matter, Intel, AMD, Nvidia, Mediatek, were still pumping out Chip with HEVC support by default after the HEVC Advance announcement. The TV industry does not seems to care, blog like Streamingmedia were still cheering for HEVC to come mainstream. VP9 or AV1 will be too late to the game. Which i agree, a specification, and a product out there to support it takes time.

Then this blog post, from developer of x265. I smell something is going on.

I agree Software Decoder should be free. It instantly removes any barrier to be included in Firefox and Chrome.

Then I think Apple's decision will likely make a big impact. Neither side has Apple on broad yet.

[hajj_3]

@x265_project is there any update yet, have companies decide to create a single patent pool with fair prices yet?

[WhatZit]

Whilst I understand the mentality behind's MPEG-LA's belief that the benefits of HEVC are both self-promoting and desirable at a grass-roots consumer level, the potential adopters of HEVC are left with a much more grounded belief: why bother?

Where are all the forums full of Android or IoS users demanding x265 NOW?

Where are all the streaming content providers placing orders for upgraded HEVC hardware boxes NOW?

They're not there for one simple reason: lack of demand. It's not (directly) money, or license hoop-jumping, it is, purely and simply, consumer demand.

Why is there no demand? Well, how many consumers even KNOW about HEVC? Percentage? I'll be optimistic and say 5% (that's pretty optimistic).

Why don't 95% of consumers know how great HEVC is? Because, it has NEVER been promoted as a wonderful "future-proof" feature/enhancement in new devices.

Why has it never been promoted by hardware manufacturers? Because they're being asked to pay for a product that has no demand! Worse, they're being asked to pay to promote someone else's product! Bingo, there it is.

Frankly, if MPEG-LA want their precious HEVC standard to be adopted, they better accept that they need to compromise on the initial promotion period. If they're not prepared to fund their own marketing, then how dare they expect anyone else to do it for them?

So, one way to accelerate PROMOTION, therefore the subsequent ADOPTION, of HEVC is to GIVE AWAY FOR FREE promotional licenses to hardware manufacturers for a limited period. This way, devices can include HEVC decoding as cheaply as possible, market it as the next big thing as cheaply as possible, and drive sales (ergo demand) as cheaply as possible.

After the limited promotional period (18 months?), THEN MPEG-LA can start to gradually raise the fees up from zero FOR NEW DEVICES.

Above all else, MPEG-LA may want 100% of their avaricious license, but 100% of 0 is 0, whereas 10% of 1000000 is 100000. Common sense, I know, but its amazing how many people don't think like this.

[iwod]

@WhatZit You are making it as if MPEG-LA is to blame, which is not entirely true because MPEG-LA is not longer in control of HEVC licensing. HEVC is to blame when they introduce their ridiculous pricing system. Halting the whole industry for a year. Now they have announced a new pricing, many companies are rethinking if it make sense. This will likely halt the industry for another year at least.

That is two whole year for Open Media to catch up on.

[CruNcher]

Whilst I understand the mentality behind's MPEG-LA's belief that the benefits of HEVC are both self-promoting and desirable at a grass-roots consumer level, the potential adopters of HEVC are left with a much more grounded belief: why bother?

Where are all the forums full of Android or IoS users demanding x265 NOW?

Where are all the streaming content providers placing orders for upgraded HEVC hardware boxes NOW?

They're not there for one simple reason: lack of demand. It's not (directly) money, or license hoop-jumping, it is, purely and simply, consumer demand.

Why is there no demand? Well, how many consumers even KNOW about HEVC? Percentage? I'll be optimistic and say 5% (that's pretty optimistic).

Why don't 95% of consumers know how great HEVC is? Because, it has NEVER been promoted as a wonderful "future-proof" feature/enhancement in new devices.

Why has it never been promoted by hardware manufacturers? Because they're being asked to pay for a product that has no demand! Worse, they're being asked to pay to promote someone else's product! Bingo, there it is.

Frankly, if MPEG-LA want their precious HEVC standard to be adopted, they better accept that they need to compromise on the initial promotion period. If they're not prepared to fund their own marketing, then how dare they expect anyone else to do it for them?

So, one way to accelerate PROMOTION, therefore the subsequent ADOPTION, of HEVC is to GIVE AWAY FOR FREE promotional licenses to hardware manufacturers for a limited period. This way, devices can include HEVC decoding as cheaply as possible, market it as the next big thing as cheaply as possible, and drive sales (ergo demand) as cheaply as possible.

After the limited promotional period (18 months?), THEN MPEG-LA can start to gradually raise the fees up from zero FOR NEW DEVICES.

Above all else, MPEG-LA may want 100% of their avaricious license, but 100% of 0 is 0, whereas 10% of 1000000 is 100000. Common sense, I know, but its amazing how many people don't think like this.

Do you actually realize that your Logic is the same like that of a Drug Dealer ;)

Frightened isn't it ;)

[x265_Project]

Do you actually realize that your Logic is the same like that of a Drug Dealer ;)

Frightened isn't it ;)
If the product was something bad and harmful, offering a free trial period would be evil. But we're talking about a new technology that can benefit everyone who produces, distributes or consumes video (or devices that support video). We've clearly got a "chicken and egg" problem right now. Which comes first; the device support for HEVC or the demand for HEVC device support from application developers? It's clear that the current licensing programs offered by MPEG LA and HEVC Advance are not succeeding on consumer compute devices, as the device OEMs aren't ready and willing to sponsor the cost of HEVC patent licensing - particularly when there are very few application developers or web services using HEVC. To succeed they need to create demand. Allowing software developers to use HEVC without charging a new royalty (that they don't pay now for AVC or any other video codec) just makes sense.

[iwod]

If the product was something bad and harmful, offering a free trial period would be evil. But we're talking about a new technology that can benefit everyone who produces, distributes or consumes video (or devices that support video). We've clearly got a "chicken and egg" problem right now. Which comes first; the device support for HEVC or the demand for HEVC device support from application developers? It's clear that the current licensing programs offered by MPEG LA and HEVC Advance are not succeeding on consumer compute devices, as the device OEMs aren't ready and willing to sponsor the cost of HEVC patent licensing - particularly when there are very few application developers or web services using HEVC. To succeed they need to create demand. Allowing software developers to use HEVC without charging a new royalty (that they don't pay now for AVC or any other video codec) just makes sense.

They are simply too greedy. As I posted in the Apple HEVC support thread, Apple simply doesn't think the cost is fair. Hence why they backdown from most HEVC support on iOS devices.

At 25M Cap from MPEG-LA, and 50M Cap from HEVC Advance, a total of 75M per year. And this doesn't even include Technicolor. If the industry adopted, that is Apple, Qualcomm, Mediatek, Intel, AMD, Nvidia, Samsung, HiSilicon, AllWinner. All these company will definitely reach the Cap, assuming it is the SoC maker who paid the fees. If you add up all the small players, that is well over 1 Billion Royalty fees per year. On a typical life time of Video Codec, they could connect 10 Billion of Royalty.

I am not sure if you need to pay royalty if you are not even operating in US. Pretty much all if China's SoC, from TV Box to players all have HEVC support already.

[WhatZit]

Do you actually realize that your Logic is the same like that of a Drug Dealer ;)


To paraphrase Maxwell Smart: "If only they'd used their marketing innovations for good, instead of evil..." Although video encoding is a somewhat addictive lifestyle :)

Here's another "drug-dealer" suggestion to foster promotion: have a royalty-free "cut-down" branch of HEVC available for commercial use that severely limits user configuration in encoding (i.e lock it at CRF26+FASTER), and restricts hardware decoding to 576P maximum or something.

Same deal as before: With the minimum of resistance/pushback, get the commercial software publishers to include HEVC in all of their encoder/editor products, and get the device manufacturers including HEVC decoding in all of their hardware.

Honestly, if I was MPEG LA (or one of those rabid splinter groups that people have been pointing out), I would be doing every bloody possible thing in my power to get it out "there", now! Right NOW!

[x265_Project]

http://www.prnewswire.com/news-releases/hevc-advance-announces-royalty-free-hevc-software-300367212.html (HEVC Advance makes software-only implementations royalty free)

[iwod]

http://www.prnewswire.com/news-releases/hevc-advance-announces-royalty-free-hevc-software-300367212.html (HEVC Advance makes software-only implementations royalty free)

I just read this else where, and OH MY GOD Hell just frozen over.

Now we just need MPEG-LA to do the same. Then we can at least turn every god damn jpeg into bpg.

And super fast software decoder will be critical.

I never thought this will come so fast. May be it is still not too late.

[sneaker_ger]

Now we just need MPEG-LA to do the same.
Have all known patent owners joined one of the two? I remember some owners hadn't some time ago. For example I don't see Fraunhofer on any of the lists.