x265/HEVC coming next year

Cool stuff. How long away ??

x264 is just one software implementation of the h264 standard, you are using its name like it is the standard itself.

I'm heavily involved in the HEVC (a.k.a h265) standardization process. It truly is a "miracle" codec as some of the detractors are calling it. The reason they detract is math. h.265 exists in conceptual form, and videos have been encoded from raw sources with it. There are several problems with ratifying this standard and none of them are purely technical:

1. Math. (Yes, this is technical but...) h.265 will require a tremendous amount of processing power on general purpose CPUs. This is no different than h.264 AVC when it came out and we were all bawling on this very forum about single core chips trying to decode 1080p60 or even 1080p24 with CoreAVC. The march of time and progress.... Still many detractors are saying it will alienate too much hardware as you'll probably need a Corei7 gen2 or better to do 1080p in real time (decode) unless some of the optimizations make it in (see point 2)
2. Concessions. H.264 SVC proponents (itself a collection of proprietary formats, the only one freely available being PLCM's just recently) want to layer video at different bitrates, especially for real-time video communications devices (the segment where my company exists, and my skin in the game) This has it's own problems. Also, several companies want h.265 to have a low-complexity mode that allows less powerful devices to decode at lower quality, which involves either lowering overall quality, or increasing payload (bitrate or bandwidth depending on your point of view)
3. Timing. Many companies are investing huge development time into AVC and SVC in the h.264 world (my company included). So, timing is everything. h.265 could have been ratified long ago if it weren't for many businesses wanting to recoup development costs, and others wanting to see those who invested in h.264 SVC fail.

tl;dr - politics and money are slowing down progress on h.265. Don't expect it until mid 2013 at the earliest, 2014 is quite possible. Hacked "draft" versions may make an appearance in 2013 a-la 802.11n's famous ridiculous history.

Good Question Tiddles88 -
Like an earlier poster said, "x265" is nonexistent (to my knowledge) and would likely be *one way* to decode h.265/HEVC. (i.e. a "filter" in our world, a decoder)
h.265 would be the codec standard's name. (For both encode and decode) and various organizations and companies could choose to decode it as they wish. Encode is the most important, with decode as a reference in the standard, when it comes to h.265 or any other video codec. Many companies try all sorts of tricks and gimmicks to decode codecs of all manner today. It's easier if we change the names:

Example:
I am going to make a new way of using lossy encoding methods to compress video. I'm going to call it Monkey Poop in. Monkey Poop In compresses raw video data or decompressed data from another codec in a very specific way. I am also releasing Monkey Poop Out, which is my preferred way of getting Monkey Poop back to the raw video needed for your display to show 1920x1080 pixels 60 times a second. I'm calling both of these methods together "Monkey Poop" and I will charge money to use either method in your product. To further make it complicated, I will charge less money to just use Monkey Poop In or just Monkey Poop Out in your product. (Think a BluRay Player - it only needs Monkey Poop Out)

Now let's say some enterprising people come along. They want to find another way to decode Monkey Poop In. (Since many more people are interested in free ways to decode than encode) We'll call this group of people FFMpeg . They reverse engineer the mathematical tricks I am using to encode 60 bitmaps of 1920x1080 images every second into much smaller files. They release "xMonkeyPoop" since they want their version of Monkey Poop Out to look as much as an official decoder as possible. They could also reverse engineer the encoding methods of Monkey Poop In to where their code is indistinguishable from Monkey Poop In, and everyone can use it. They could be really confusing and call this same method "xMonkey Poop" So now we would have:

Monkey Poop: A pair of Encoder and Decoder software mechanisms (Monkey Poop In and Monkey Poop Out) that is commercial. It costs you money to use this in your product.
xMonkey Poop: A pair of Encoder and Decoder software mechanisms (xMonkey Poop refers to both directions) that is free (or free-ish) to use in your product.

Example over.

h.265 is an effort to make a real standard that we can all conform to. It would be non-commercial but many people want it to be commercial. Decode, like any lossy algorithm, would be much less compute-intensive than encode (finally getting to your question) but the methods employed by the proposed h.265 algorithm(s) are extremely compute intensive on both sides of the equation, especially due to some advanced mathematics involved at very high frequencies which were not possible even 5 years ago on common hardware. Specifically things like "dot product" and DiffEq on a CPU were very expensive in terms of cycles and better suited to GPUs. The decode will still need to do those complex parallel operations. Perhaps the advent of this "APU" will make things more common, but there is also a huge push to mobile and ARM which is slowing down the whole concept, but is important for wide user adoption.

I'd like to hear your opinion on how you think this is going to shake out. What are the chances we will get a truly free to implement h.265 standard versus something patent encumbered, even if it is FRAND (aka fair, reasonable and non-discriminatory, except to discriminate against people who want to give their implementation away for no money)?

The h.264 FRAND status has caused a bunch of headaches for Firefox who can't afford to pay a "fair" licensing fee since they give their software away for no money and not all of their supported platforms can be relied on to have an h.264 implementation present. It would be nice if h.265 fixed that problem too.

FYI: https://code.google.com/p/x265/ -- although the author just suspended public work yesterday, he does seem to have an association with the x264 project so I think his use of the name is at least semi-legit.

This project is not related to the x264 encoder, and not condoned by them.

Regarding the "enterprising people", i know that work on a decoder is already under way for FFmpeg, and other developers are also looking into encoders (not related to the "x265" project mentioned above)

PS:
Its not reverse engineering if you have a standard/spec to use for implementation. A true reverse engineering of a codec as complex as HEVC would be a rather challenging task.

HEVC certainly looks good and it will be the first major video standard that supports a 10-bit profile aimed at consumers which will hopefully mean that 10-bit video will be seen in consumer video formats (such as the 4K version of Blu-ray).

A recent proposal at the October HEVC meeting showed that a single core of a 1.3 GHz ARMv7 CPU could decode a 2 Mbps 1080p30 HEVC video. Also in the long term smartphones will use HEVC hardware decoders.

The low complexity version of HEVC was removed from the HEVC draft over a year ago. There are a few SVC proposals for HEVC that use a low complexity base layer but they have nothing to do with the first version of HEVC.

SVC makes sense for video conferencing but it has limited application outside of that area. Hopefully the Main 10 profile of HEVC will be used in consumer products.

I am very skeptical of that. From everything I have read the development of HEVC is on schedule. The companies working on HEVC as well as the official HEVC meeting documents state that the final draft of HEVC will be released in January 2013. That will be the first version of HEVC and additional HEVC profiles will be added in the future.

So can I ask a noob question... what's better about 265 that someone would want it ?

In the simplest terms, achieve a 50% reduction in file size for the same visual quality at the cost of more computational power.

Here are the three reasons I would give for why HEVC will be better:

1. HEVC will be at least twice as efficient as MPEG-4 AVC for video compression.
2. HEVC will allow for 10-bit consumer video. MPEG-4 AVC has a High 10 profile that allows for 10-bit video but it is a professional profile that has 3x the maximum bit rate of the 8-bit Main profile and as such the 10-bit profile has never been supported by consumer hardware decoders. The 10-bit profile in HEVC was specifically designed for consumers and has the same maximum bit rate as the 8-bit profile. Hopefully the Main 10 profile in HEVC will become standard in consumer hardware and consumer video formats. 10-bit video has long existed in the professional world but there is now a chance that it will be seen in the consumer world.
3. HEVC is designed for 4K and 8K resolutions. MPEG-4 AVC supported 4K but it was aimed at the professional world. For example what if a company next year wanted to make a consumer video player capable of 4K at 24 fps with 10-bit video? With MPEG-4 AVC the hardware decoder would need to support a maximum bit rate of 720 Mbps. With HEVC the hardware decoder would need to support a maximum bit rate of 25 Mbps (with an option of 100 Mbps). HEVC is designed with the idea that 4K will be used in consumer products.

Edited by Richard Paul - 12/4/12 at 4:17pm

You can do a 10-bit encode with x264 which will actually be smaller in size at the same quality than a corresponding 8-bit encode, your 3x bit rate argument is rather bogus.

The improved efficiency will also come with a cost, in hardware requirements. The computational complexity required to get that high efficiency increase will require a lot of computional power, mostly during encoding, but also a big increase during decoding.
So even if the required bitrate is less, the power required for decoding certainly will not go down.

PS:
I don't want to see 4K at 25Mbps, not with HEVC or any other codec. Even for HEVC thats just not enough bits to be any good.
Edited by Nevcairiel - 12/3/12 at 5:33am

What consumer products with hardware decoders support the 10-bit profile of MPEG-4 AVC? I have not heard of one. Excluding computers and game consoles there is a good reason that video decoding is almost always done with hardware decoding. Hardware decoding is cheaper and uses much less power than software decoding which is why Blu-ray players, DVD players, cable/satellite STBs, and streaming STBs (Netflix, Vudu, etc...) use hardware decoders. If a company wants to make a hardware decoder they have to follow the standard. That is good since it creates compliant products but it is also means that if the standard requires a very high bit rate than the hardware decoder must support that bit rate. The 10-bit profile of MPEG-4 AVC was designed with a bit rate aimed at the professional world which is why the maximum bit rate is 3x higher. That makes sense for professional products but it makes the hardware decoders more expensive. That is one of the reasons why the 10-bit profile of MPEG-4 AVC has not been supported by consumer hardware decoders.

That will likely be a matter of debate in the future but I would mention that most 1080p24 video streaming services are done at 5 Mbps or less. If you were to increase that by 4x for 2160p24 streaming that would be 20 Mbps. From what I have read HEVC is about 3x more efficient at 4K resolution than MPEG-4 AVC. As such to have the same video quality as most video streaming services HEVC would need a bit rate of about 7 Mbps for 4K at 24 fps.

In my opinion the Main tier of HEVC makes sense for cable, satellite, and streaming. A maximum bit rate of 25 Mbps for 4K at 30 fps should be sufficient. A maximum bit rate of 40 Mbps for 4K at 60 fps should be sufficient. Many of the companies who make video decoders or who are involved with broadcasting are the ones that created the two tiers. As such from what I have read the Main tier will be sufficient for the vast majority of video services.

Good job avoiding the bit rate argument entirely. Did i say anything about hardware support? No? Right!

I personally still doubt that the broadcasters will use the 10-bit mode in HEVC, its hard to market that as people don't fully understand it.

"Should be"? I think broadcasts look terrible (especially because they are even 1080i most of the time), and you lose about zero quality when you just compress a 1080p broadcast to 720p because of the low bitrates. So no, it should not be sufficient, that the broadcasters are too cheap to give us real quality is another problem entirely. Hopefully we won't kill optical media before we have real alternatives.
Blu-rays allow a bit rate of up to 40Mbps with H.264, and i certainly hope we won't see 4k material on optical discs below 100Mbps with HEVC. Luckily they also defined profiles and levels far higher then your 25Mbps, but sadly broadcasters always cheap out. They advertise with 4K and then shove the lowest possible bitrate down your throat, how fun.

PS:
AFAIK, US broadcasts are mostly MPEG2 even, usually at 10-15Mbps, so the comparison is even harder to make.
Edited by Nevcairiel - 12/3/12 at 11:12pm

In the post you responded to I was referring to the maximum bit rate for the 10-bit profile of MPEG-4 AVC. I thought that was clear in my post since I was referring to the the bit rate of the profile. I used the term "maximum bit rate" a few times in the next paragraph but I now notice that I didn't use it in the paragraph that you quoted. Just to make this clear I was referring to the maximum bit rate of the High 10 profile of MPEG-4 AVC.

A higher bit depth is harder to explain than higher resolution but it is one of the main improvements that could be made with consumer video. Also I think that a higher bit depth will be needed for the UHDTV color space which is a lot larger than the HDTV color space.

Well only time will prove which one of us is right but I think that good video quality will be possible with HEVC using the Main tier.

I think this discussion cements why I want to get a much more powerful CPU in my next PC. The rule on the forums is that you don't need quad core, the modern i3 is powerful enough. If I bought and i3 Ivy Bridge right now it doesn't sound like that will be enough to handle this once implemented.

My htpc is five years old now, which I know in tech is ancient, but it's too big an investment to buy another one every two or three years. Compressing a movie using h264 takes around 36 hours on my system to get a 4-8gb file. That's 36 hours that my computer is unusable for other tasks - so I don't do it. I still use xvid when I need to compress something.

I think there is some hurdles before you see it's wide spread implementation.

So buy a cheaper system, then it won't be such a big investment and you can do it more frequently.

Agreed, I think we might see some software based services using HEVC next year but it most likely won't be until 2014 that we see consumer products that have mass produced HEVC hardware decoders. Patents could also be a potential issue depending on how things go though the MPEG LA is trying to get all the companies involved in their patent pool. Many people dislike the MPEG LA but a common patent pool is a lot better than having several dozen companies trying to collect royalties.

well that basically means three things, cpu, motherboard, and ram. and thats going to be at least $200.

one of the htpc design goals always seems to be low power use. and if you're going to need to do encoding/transcoding, its not really an htpc anymore and you'd be better served with another computer. ok, so now im building two computers, running both of them all the time, one to record television and watch movies, one to compress for storage and run plex to mobile devices and google tv.

i just feel like so many people follow a "minimum cost and power to achieve my current goal" and consideration isnt put into the future. id like to run SVP, to transcode DTS for my Logitech Revue, to power both extenders, to compress video on the fly to send it over the web. i dont want one thing to suffer because somebody else is trying to do another.

had i spent a bit more money on a quad core five years ago I might be in a bit better position now than having to throw away all my HBO recordings away and buy an Ivy Bridge/motherboard/ram.

Buy a quad-core now and put it in your current motherboard. That's exactly what I did 9 months ago with my dual-core htpc from 2007. They are a lot cheaper today than they were back then too. Bleeding edge always has a premium, so unless you need the functionality today, it is a waste of money to pay today for performance you only expect to need 2 years down the road.

ITU approves the H.265 video format:
http://www.engadget.com/2013/01/27/itu-approves-h265-video-format/

Cool stuff

so basically all my media is going to be obsolete or taking up too much space ????

and i can't stream it effectively.... as new H265

It's no more "out of date" than avi/xvid/divx/whatever from yesteryear. They will still play. They will still look the same. Aside from less disk space, I don't see a huge immediate advantage to re-encode all your media ASAP if ever.

+1

Totally agree and what ever hardware is released to handle this new encoding I am sure it will be backwards compatible to current popular codecs (ie h264, xvid, mpeg etc.).