x265 - Work already started?

[iwod]

http://code.google.com/p/x265/

https://github.com/chenm001/thevc

I was wondering why it isn't based or branch off from x264. Are they starting from scratch?

[poisondeathray]

There is a blog link on the google page, but it's in chinese

http://chenm003.blog.163.com/

There seems to be only 1 author/contributor ? Where are the other "usual suspects" ?

[LoRd_MuldeR]

Doesn't look like the "x265" project you are linking to is created by the people who created x264.

Also it's under an BSD license, which means they can't take any code from x264, which is under the GPL:
https://github.com/chenm001/thevc/blob/master/COPYING

(There is no "Copyleft" in the BSD license, which is an important idea of the GPL, so they are not compatible)


Finally we need some prove for those claims:
Simple and powerful HEVC/H.265 implement.
Optimize for embedded, FPGA, GPU, MultiCores system.

Given that h.265/HEVC has not even been officially released (scheduled for January 2013), I doubt their implementation is complete or even optimized yet ;)

(Maybe that implementation is based on an early draft version of h.265/HEVC)

[poisondeathray]

Wouldn't it be a better idea to join forces ??

What are the x264 dev's thoughts ?

[LoRd_MuldeR]

Update: Actually it appears that "x265" and "thevc" may be two different projects, just the x265 author is working on the latter too.

(So forget about my comment about the license issue, as x265 actually is (http://code.google.com/p/x265/source/browse/COPYING) GPL'd)

[hajj_3]

think the final draft of HEVC/H265 is out next month.

[burfadel]

Its kind of cheeky and a little deceiving to be using the name x265 if the projects are completely unrelated, because it would be human nature to expect it to be from the same authors and to the same quality as x264...

h265 is a long way off yet, as its target is a 50 percent improvement in the compression vs size. This is achievable, but requires large amounts of CPU. Also no point if the standard encode settings (and use for transmission) of h265 isn't much better than h264 for the same CPU usage. If you encode on tight x264 settings and the resultant size is say, 500MB, and you compress in h265 to the same CPU usage (may be low settings) and for the same quality its 500MB or larger, its a bit of a disadvantage.

That aside, to get 50 percent better than h264 it may encode at 0.17 fps compared to the 100 percent larger h264 (using x264) encoding at 25fps...

A note on the percentages above, seeing people online often misinterpret or incorrectly state percentages:
h265 half the size, is 50 percent better. Since its half h264, you need to double it to get back to h264 size, meaning 100 percent larger.

Processing requirements of h265 for its target size etc is one reason why it hasn't been released yet. Little point releasing something unusable that can be possibly improved if delayed :). h265 is probably the logical choice for future Ultra High Definition Television (UHDTV) transmission and distributable media (blu-ray etc), since its 7680×4320. It would also be the logical choice for Quad Full High Definition (QFHD) 3840×2160 if thats the future resolution instead of UHDTV.

Probably wouldn't have been a bad idea for these people to work on x264 and use what they can from it once the standard is released. At the moment, they could do a whole heap of work on h265 only to find they have to redo the whole lot if the standard changes before releasing :S

[hajj_3]

i'm sure fixed function hardware will make h265 fast to encode, just look at the speed of quicksync encoding and ati's upcoming equivalent. I'm sure ARM chips will be able to do it too but existing chips will indeed be very slow i'm sure. Hopefully existing hardware decode cpu/apu's might be able to add support to hardware decode h265 but can't see them being able to hardware encode h265.

[nm]

Hopefully existing hardware decode cpu/apu's might be able to add support to hardware decode h265
Not possible with current hardware. Most likely it will take a couple years before hw decoders start popping up in GPUs and other generic consumer devices. The format needs to be adopted widely first.

[iwod]

Well we can do the decoding with special Hardware. All Mobile Devices will force this to happen compared to previously they all try to use the CPU as much as possible.

Since UHDTV will have 4 times the pixel compare to 1080P Full HD, i wonder would 50% reduction be enough?

[Ghitulescu]

What is the goal of H.265?

[LoRd_MuldeR]

From Wikipedia:

Said to improve video quality and double the data compression ratio compared to H.264., HEVC can scale from 320 x 240 pixels all the way up to 7680 x 4320 resolution.HEVC aims to substantially improve coding efficiency compared to AVC High Profile, i.e. to reduce bitrate requirements by half with comparable image quality, at the expense of increased computational complexity. Depending on the application requirements, HEVC should be able to trade off computational complexity, compression rate, robustness to errors and processing delay time.

We'll see how much of that will hold true ;)

And we probably won't know until an H.265 encoder with the same level of optimizations as x264 has nowadays will be available - which can take years!

(Remember: There are enough "bad" H.264 encoders, which could easily lead to the conclusion the H.264 is worse than MPEG-4 ASP, if there weren't the "good" ones to prove the opposite)

[Ghitulescu]

If no "heavy" stays behind, the fate of H.265 will look alike the one of on6 codec, or any ogg ones. :(

[mandarinka]

Its kind of cheeky and a little deceiving to be using the name x265 if the projects are completely unrelated[...]

Note that x264 just applies an x letter to "h.264", most likely inspired by xvid. So continuing the tradition with x + h.265 is not particularly cheeky. It's an ambuitious name (much like x264 at the start when xvid was the choice codec), but that's it.

[Dark Shikari]

There seems to be a history of projects like this. For example, there was a Chinese developer who has been (for years) spam-mailing hundreds of video codec developers with his "H.265" codec which has absolutely nothing to do with the actual H.265.

[poisondeathray]

x265.nl , x265.com aren't registered domains yet . Maybe someone should register before someone makes a play

[mariush]

Well, I registered x265.net now.

If Dark Shikari or one of the x264 guys want to start on x265 sometimes in the future, I'll be willing to transfer it to them for free.

[iwod]

There seems to be a history of projects like this. For example, there was a Chinese developer who has been (for years) spam-mailing hundreds of video codec developers with his "H.265" codec which has absolutely nothing to do with the actual H.265.

So would we get an official x265 project? Since there is now even a x262

[Dark Shikari]

H.265 is quite different from H.264 in terms of structure; it'd be a pretty hefty modification.

But if someone really wants to try...

[hajj_3]

i think the only real advantage to most people is smaller filesize for the same quality, 700mb files would be far smaller. I can't see that many people using this for the higher quality as it would take ages to encode i'm sure. Needless to say i'd love to see an x265 real project :)

[CruNcher]

i would say from a consumer standpoint we can be pretty happy with what H.264 already accomplished in my terms it was the major first step in the right direction (though it also showed its "dark" side pretty obvious, and yes that's equivocally meant *wink* ) and many will follow though i don't see H.265 there (and don't feel like hyping around it) just for the sake of it but i guess it matters from your standpoint ;)

[shon3i]

@Dark Shikari, what happend with this (http://forum.doom9.org/showthread.php?t=154430) ? Can we see some results since there some H265 encoder.

[iwod]

i would say from a consumer standpoint we can be pretty happy with what H.264 already accomplished in my terms it was the major first step in the right direction (though it also showed its "dark" side pretty obvious, and yes that's equivocally meant *wink* ) and many will follow though i don't see H.265 there (and don't feel like hyping around it) just for the sake of it but i guess it matters from your standpoint ;)

It is merely the thought of having another standard to replace the Main Profile which is currently limited by Apple Devices. Since it is extremely unlikely we will get High Profile supported.

Also, starting ASAP, would means we can wait less before the implementations mature. It was merely my naive thoughts that extending x264 to support H.265 would be much faster, and like everything in life, easier said then done.

[nm]

It is merely the thought of having another standard to replace the Main Profile which is currently limited by Apple Devices. Since it is extremely unlikely we will get High Profile supported.
Both iPhone (starting with 3GS) and iPad 1 & 2 support High Profile H.264.

[amtm]

Though officially high profile support is only stated in the iPhone 4S and iPad 2 tech specs. The 3GS tech specs states baseline and the iPhone 4 and original iPad tech specs states main profile. Obviously whether they can support it contrary to the official specs is another matter.

[iwod]

Both iPhone (starting with 3GS) and iPad 1 & 2 support High Profile H.264.

Can anyone confirm that? I try a few google search and nothing meaningful turned up.

[LoRd_MuldeR]

That's what Apple says about the 4S:

Unterstützte Videoformate: H.264 Video: bis zu 1080p, 30 Bilder pro Sekunde, Main Level Profile Level 4.1 mit AAC-LC Audio bis zu 160 KBit/s, 48 kHz, Stereo-Audio in den Formaten .m4v, .mp4 und .mov; MPEG-4 Video: bis zu 2,5 MBit/s, 640 x 480 Pixel, 30 Bilder pro Sekunde, Simple Profile mit AAC-LC Audio bis zu 160 KBit/s, 48 kHz, Stereo-Audio in den Formaten .m4v, .mp4 und .mov; Motion JPEG (M-JPEG): bis zu 35 MBit/s, 1280 x 720 Pixel, 30 Bilder pro Sekunde, Audio im Format ulaw, PCM Stereo-Audio im Format .avi

Source: http://www.apple.com/de/iphone/specs.html

Though this doesn't mean it won't accept more than that in reality. They just don't guarantee more ;)

[JEEB]

Can anyone confirm that? I try a few google search and nothing meaningful turned up.
I can confirm this. iTunes used to stop you from transferring the files over, but if you got the files in using some other way, they would play on 3GS or newer (high profile, up to either level 4 or level 4.1).

[nm]

Can anyone confirm that? I try a few google search and nothing meaningful turned up.

What would be meaningful? Apple doesn't announce the support officially in their spec sheets, so you'll only have the word of mouth. But it's common knowledge by now:

http://forum.doom9.org/showthread.php?p=1330324#post1330324
http://forum.doom9.org/showthread.php?p=1453029#post1453029


That's what Apple says about the 4S:

Huh, the deutsch page has an error. They do claim High Profile support for 4S in other languages. From the english page (http://www.apple.com/iphone/specs.html):
Video formats supported: H.264 video up to 1080p, 30 frames per second, High Profile level 4.1 with AAC-LC audio up to 160 Kbps

[amtm]

That's what Apple says about the 4S:



Source: http://www.apple.com/de/iphone/specs.html

Though this doesn't mean it won't accept more than that in reality. They just don't guarantee more ;)

Then they haven't updated their German page. From the US page (http://www.apple.com/iphone/specs.html) about the iPhone 4S they list:

Video formats supported: H.264 video up to 1080p, 30 frames per second, High Profile level 4.1 with AAC-LC audio up to 160 Kbps,

[CruNcher]

Hehe interesting though wouldn't surprise me if they came up with the idea to make it a US Firmware feature only (there where already so crazy ideas in restriction stuff see Sonys escapade of device restrictions), though most likely someone of the Webmasters f***ed that up ;)

Hehe how funny is that http://www.apple.com/fr/iphone/specs.html the french only get Baseline sorry guys ;)
http://www.apple.com/de/iphone/specs.html <- Germany Main Profile (we contributed so much to H.264 and now we only get Main Profile screw you Apple, especially "Main Level Profile Level 4.1" whatever that is ;))
http://www.apple.com/nl/iphone/specs.html <- Netherlands High Profile
http://www.apple.com/no/iphone/specs.html <- Norway High Profile
http://www.apple.com/se/iphone/specs.html <- Sweden High Profile
http://www.apple.com/fi/iphone/specs.html <- Finland High Profile
http://www.apple.com/ru/iphone/specs.html <- Russia High Profile
http://www.apple.com/iphone/specs.html <- US High Profile
http://www.apple.com/ca/iphone/specs.html <- Canada High Profile
http://www.apple.com/uk/iphone/specs.html <- UK High Profile
http://www.apple.com.cn/iphone/specs.html <- China High Profile
http://www.apple.com/tw/iphone/specs.html <- Taiwan High Profile
http://www.apple.com/jp/iphone/specs.html <- Japan High Profile
http://www.apple.com/kr/iphone/specs.html <- Korea High Profile
http://www.apple.com/es/iphone/specs.html <- Spain High Profile

Anyway took Apple long enough to support High Profile officially @ all even if it was introduced silently with the 3GS already

[amtm]

Hehe interesting though wouldn't surprise me if they came up with the idea to make it a US Firmware feature only

Except it makes no sense? What would they gain by restricting German phones to only main profile other than huge headaches from having to maintain separate firmwares? It is sloppy updating and nothing more.

[CruNcher]

You don't need to maintain separate firmwares you could do it automatically based on the id of the current firmware, though as i said this is most probably really a webmaster fault especially the French Webmaster should be ashamed of himself ;)

[Kurtnoise]

Hehe how funny is that http://www.apple.com/fr/iphone/specs.html the french only get Baseline sorry guys ;)
where did you read this ?

Formats vidéo pris en charge : vidéo H.264 jusqu’à 1080p, 30 images par seconde, profil de référence jusqu’au niveau 4.1 avec son au format AAC-LC jusqu’à 160 kbit/s
==
Video formats supported: H.264 video up to 1080p, 30 frames per second, High Profile level 4.1 with AAC-LC audio up to 160 Kbps,

[CruNcher]

http://translate.google.com/#auto|en|Formats%20vid%C3%A9o%20pris%20en%20charge%20%3A%20vid%C3%A9o%20H.264%20jusqu%E2%80%99%C3%A0%201080p%2C%2030%20images%20par%20seconde%2C%20profil%20de%20r%C3%A9f%C3%A9rence%20jusqu%E2%80%99au%20niveau%204.1%20avec%20son%20au%20format%20AAC-LC%20jusqu%E2%80%99%C3%A0%20160%20kbit%2Fs

interesting it doesn't references anything in the text indeed i wonder where the algorithm or the Google Watson catched Baseline or Main Profile from and why :D

though for me "profile de reference" sounds like reference profile or another word to describe reference with would be the base nothing like High Profile http://translate.google.com/#auto|en|profil%20de%20r%C3%A9f%C3%A9rence i guess that's also what Googles Watson combines :D

shouldn't it be


Formats vidéo pris en charge : vidéo H.264 jusqu’à 1080p, 30 images par seconde, haut profil jusqu’au niveau 4.1 avec son au format AAC-LC jusqu’à 160 kbit/s

or


Formats vidéo pris en charge : vidéo H.264 jusqu’à 1080p, 30 images par seconde, profil de haut jusqu’au niveau 4.1 avec son au format AAC-LC jusqu’à 160 kbit/s

[PhrostByte]

A note on the percentages above, seeing people online often misinterpret or incorrectly state percentages:
h265 half the size, is 50 percent better. Since its half h264, you need to double it to get back to h264 size, meaning 100 percent larger.

Half the size does not mean 50 percent better. Does H.265 aim to give 50% more quality per bit, or files which use 50% the bits for the same quality? It makes a big difference.

50% better quality = 66% the size. (ie. 1400MB H.264 -> 933MB H.265)
100% better quality = 50% the size. (ie. 1400MB H.264 -> 700MB H.265)

[hajj_3]

it looks like some good progress on HEVC/H.265: http://www.h265.net/2012/02/final-meeting-notes-for-1st-to-7th-jct-vc-and-working-drafts.html

[iwod]

it looks like some good progress on HEVC/H.265: http://www.h265.net/2012/02/final-meeting-notes-for-1st-to-7th-jct-vc-and-working-drafts.html

Finally a update after a year of silence.

[Shevach]

I attended at two last HEVC meetings and have been monitoring HEVC activity since the project started.
Moreover I manage two threads on HEVC in LinkedIn.

HEVC is potentially better than H.264. Subjective tests ( were conducted by NTT DOCOMO, Inc) between HEVC (HM 5.0) and JM 18.2 shows that absolute majority of observers find HM5.0 at half the bit-rate to be comparable or better JM 18.2 .

There are reasons to believe that HEVC is significantly better than H.264.

1) 16x16, 32x32 transforms enable much better compression of large smooth regions.

2) SAO (Sample Adaptive Offset) - applied after deblocking, minimize "distance" between reconstructed and original frame.

3) Enhanced Intra Prediction - 32 angular modes, plus DC, Planar.

4) Enhancement of MV prediction (motion vector in HEVC is predicted from four spatial neighbors (top, top-right, left, left-bottom) and one co-located MV)

5) Imrovement of CABAC

[Atak_Snajpera]

I would like to see comparision between x264 --placebo and HEVC

[LoRd_MuldeR]

I would like to see comparision between x264 --placebo and HEVC

The problem is that there is not "the" HEVC. The same way there is not "the" H.264/AVC.

For a fair comparison we would need a HEVC encoder that is as optimized as x264 is nowadays (or at least somewhat close to that).

It will probably be easy to show that current x264 can beat a run-of-the-mill HEVC encoder...

[Shevach]

I would like to see comparision between x264 --placebo and HEVC

Rate Control in HM has not been implemented yet.

[Shevach]

I believe that HEVC is better than H264.
But how many "better" or what's HEVC (HM) gain against
H264? There are many figures (e.g. in JCTVC-H0116) but HM is compared against JM.

[gyth]

I would like to see comparision between x264 --placebo and HEVC
An easier test to arrange would be to compare x264 to JM 18.2 (the current version is listed as 18.3 (http://iphome.hhi.de/suehring/tml/)...?).
Not really a fair test, but could be interesting anyway.

[poisondeathray]

There are many figures (e.g. in JCTVC-H0116) but HM is compared against JM.

Wow, looks impressive so far

http://phenix.it-sudparis.eu/jct/doc_end_user/current_document.php?id=4417

[iwod]

i was hopping many of the work on x264 could be used on x265, much like how x262 was born.

[IgorC]

From that link H.265 really compresses 2x more than H.264. http://phenix.it-sudparis.eu/jct/doc_end_user/documents/8_San%20Jose/wg11/JCTVC-H0116-v3.zip
Just in time there is a new audio format, a successor of HE-AAC/AAC. USAC.
USAC verification test report (http://mpeg.chiariglione.org/working_documents/mpeg-d/usac/USAC_vt.zip)

[hajj_3]

never heard of USAC, nice to see a better codec to HE-AAC though :)

The cost of AAC is a fortune to license i am told, multiple times more than the cost of licensing h.264. Lets hope that this will force a price reduction on aac then.

[amtm]

From that link H.265 really compresses 2x more than H.264. http://phenix.it-sudparis.eu/jct/doc_end_user/documents/8_San%20Jose/wg11/JCTVC-H0116-v3.zip

No, it compresses 2x more than the highly unoptimized JM encoder. JM hardly represents the highest compressibility possible for an H.264 encoder.

[IgorC]

Neither current reference h.265 encoder is optimal. So it's valid to compared two reference encoders to get a rough idea.


Lets hope that this will force a price reduction on aac then.
I wouldn't take a breath as licensing of MP3 doesn't get any cheap.