First Look: H.264 and VP8 Compared

[HawkMan]

  On 22/05/2010 at 19:03, boogerjones said:

I wouldn't call it a great comparison. The guy is biased since he's develops an h.264 encoder (and a good one; I use it all the time). He makes some knowledgeable code comparisons, but then compares video without telling us the settings used and concludes h.264 is magnitudes better (what a surprise from a guy with a vested interest in the codec). Later he steps out of bounds and makes some wild comments on the legal situation (which he knows nothing about).

Comparing individual frames is far less useful than comparing fluid video. And I'm not sure why everyone here is hating on the VP8 video. The obvious conclusion is that the video looks comparable at similar bitrates in the article's test. Do any of you idiots think you could pick out which video came from which codec in a blinded test? It's similar to people arguing about 128 versus 320 kbps audio; nobody can tell the difference when put to the test.

Do I think h.264 is overall a technically better codec? Yes, but we don't need the best technical codec. We need a sufficient codec free of the financial threat.

umm Fraunhofer encoded FOR him

[Danrarbc]

  On 22/05/2010 at 17:43, Elliott said:

Anyway, VP8 is looking pretty good (still inferior to H.264 as far as I can tell), but I'm holding off for performance tests (especially on mobile devices). If it's going to suck away at the battery life and CPU power of devices that currently exist, then it's not practical.

Unfortunately we already know that thanks to some odd choices On2 made it requires more CPU power to decode than h.264 Base profile. And even more unfortunately Google has already finalized things so some of those bad choices cannot be corrected.

[ichi]

  On 22/05/2010 at 19:19, Danrarbc said:

Unfortunately we already know that thanks to some odd choices On2 made it requires more CPU power to decode than h.264 Base profile. And even more unfortunately Google has already finalized things so some of those bad choices cannot be corrected.

They have frozen the specs, but that doesn't stop anyone from putting more work on the decoder.

[Danrarbc]

  On 22/05/2010 at 20:14, ichi said:

They have frozen the specs, but that doesn't stop anyone from putting more work on the decoder.

I considered saying that decoder work can still be done. But the truth is no amount of work can get around the inefficiencies in some of the encoding techniques inherent to the codec, trying to use shortcuts for instance could have compatibility consequences.

Don't get me wrong, it's great that we have a codec that is more modern here, but it's still got it's weaknesses even if it is better than Theora.

[ObiWanToby]

I for one just hope h.264 remains the standard. OSX and Windows both decode h.264 with GPU acceleration using their native players. Hell even flash claims to do the same.

The Zune and iPod also support h.264 acceleration and in the case of the Zune HD 720P output!

[PyX]

  On 22/05/2010 at 16:46, Boz said:

There is no Flash vs HTML5. They will both exist just fine as Flash works on top of HTML5.

I know what you?re saying, but trust me, there is. A lot of websites put their Flash players to garbage and roll out brand new HTML5 players, so there are market shares in that game. There is a competition.

[Elliott]

  On 22/05/2010 at 17:48, HawkMan said:

I love revisionist history. The record company that actually first removed DRM did it without apple pushing anything, in fact apple wanted them to keep DRM. they decided to drop it anyway, and then eventually the others followed. It was funny at the time to see that after apple couldn't change their mind, then they suddenly advertised how "they" had forced them to drop drm and how itunes now had DRM less music :facepalm:

Can you back up these claims at all? As far as I can tell, though I may be a little off, no record label in the US was doing away with DRM before Job's "Thoughts on Music" piece. After that, Apple and EMI announced their no-DRM partnership, and then shortly after that, Amazon put up its own no-DRM store.

  On 22/05/2010 at 17:48, HawkMan said:

The whole point was the DRM which broke the AAC spec though. instead they could have gone with the just as safe plays for sure drm that not only worked but alsop had cross compatibility and was part of the spec for that audio format. But then they couldn't lock you in so...

Yes, I'm sure they would've loved to have licensed Microsoft's now-dead (and specifically for WMA anyway as far as I can tell) DRM.

[HawkMan]

No EMI announced they where dropping DRM before the apple thing, in fact the Amazon store was supposed to be the first one. But it took some time for all the deals to finish up and Apple managed to snag the first DRM free store status, and then they rode the whole "we made them drop DRM" train to town.

It was all a bunch of theatrical bull and the apple crowd swallowed it whole despite the well known fact EMI had announced it long before the Apple supposedly made them.

Also maybe they should have licensed it, PFS remained unbroken longer than "FairPlay", and ironically it was a lot more fair. for all parties.

[Udedenkz]

Before working Flash 10.1, I had to download the video and play it in MPC:HC - otherwise it would lagg like crazy. I hope that I won't have to do that again. :crazy:

  On 22/05/2010 at 16:41, Boz said:

Actually VP8 is better in quite a few cases than H.264 Base.. it only falls behind on H.264 Main and High Profile, but since this is the codec for the web, it doesn't have to beat Main or HIgh Profile. It just needs to be as good as H.264 Base and be open source and royalty free.

It's also absolutely fantastic that they are using Matroska (probably the best container currently) and open source Vorbis.

It truly frees up the web from proprietary standards.

This statement is very BIASED toward VP8, lets examine it,

1. You separate H264 into profiles, yet you do not separate VP8 into different profiles or something equivalent.

2. You acknowledge that h264 wins using Main profile AND high profile, yet you seem to dismiss this. That is, the web stands still, and there will be no stream of medium / high bitrate 720/1080 +/-3D in the future. You give no guarantees that VP8 is future proof while you acknowledge that h264 IS. This opens up another of the many heels of VP8 - in the future there will be requirement for a new encoder / decoder standard which surpasses VP8. This implies further future hardware changes, incompatibility, inability to playback video, forced upgrades, money and time wasted (as it will be the case when switching to VP8).

3. Open source does not imply good, stop using it as if it does.

[Boz]

  On 22/05/2010 at 17:25, HawkMan said:

So we're supposed to choose open source because, not because it's better, but it's good enough... oh and it makes you feel warm and fuzzy inside... and it suffers the same patent and license probelms as the format it replaces.

yeah great reasoning. with the similarities you'd be mad to think MPEG-LA isn't going to go after them, sue, and win. and then it doesn't matter if it's open or released freely, content deliverers still need to pay so it'd be the same as h.264. only you know, not quite as good.

No.. nobody is telling you not to use H.264. It will be used for higher quality video without a doubt and it will still be used for Blu-ray and similar. The point here is that VP8 should be the bottom line of the web standard for video that everyone should accept and as such it is absolutely imperative that we have open source standard without any royalty fees. It needs to belong to people, it needs to belong to the internet. Having great open source standard for video on the web will force proprietary formats/codecs such as H.264 to work even hard to push the boundary and provide better technology.

VP8 is as good as H.264 Baseline and it will even improve. I'd say that encoder for H.264 in that example is probably much better than the test VP8 one. So VP8 encoders will improve drastically and will actually show that it's better than already mature H.264 Baseline.

In terms of patents and code, similarities do not equal win. You can argue and argue and try to blow MPEG-LA horn but the fact is that we've seen the exact same thing happening with VP3 and Theora and nothing came out of it.

  Quote

and it suffers the same patent and license probelms as the format it replaces.

No it doesn't' suffer from patent and license problems.. it's OPEN SOURCE. You are maybe hoping it will based on threats from MPEG-LA.. but it's purely speculative whether or not MPEG-LA actually even has a case here and at this point mostly fear mongering by those with vested interest in it with hopes to try to stop VP8.

  Quote

yeah great reasoning. with the similarities you'd be mad to think MPEG-LA isn't going to go after them, sue, and win. and then it doesn't matter if it's open or released freely, content deliverers still need to pay so it'd be the same as h.264. only you know, not quite as good.

All your personal opinion and really has no connection with the reality of the situation, sorry. If/Would/Should is all irrelevant. We live in the now and right now VP8 is open source and WebM KICKS BUTT as a web standard. We can discuss patents and licenses of VP8 when we actually get to that point, which to be honest, I think will never happen.

[Luis Mazza]

We don't need another video format if it's not better than H.264 with the same bitrate or with a lower bitrate and same quality just because of patents. All Windows, MAC OS an others have free options to decode H.264, so why bother with another format because it is patent-free? TODAY this is useless... Google: you should have done that back in 2004, at least.

[ichi]

  On 22/05/2010 at 23:36, Udedenkz said:

1. You separate H264 into profiles, yet you do not separate VP8 into different profiles or something equivalent.

It's the h264 profile used on the web vs the VP8 profile (are there VP8 profiles anyway?) used on the web.

  On 22/05/2010 at 23:36, Udedenkz said:

2. You acknowledge that h264 wins using Main profile AND high profile, yet you seem to dismiss this. That is, the web stands still, and there will be no stream of medium / high bitrate 720/1080 +/-3D in the future. You give no guarantees that VP8 is future proof while you acknowledge that h264 IS. This opens up another of the many heels of VP8 - in the future there will be requirement for a new encoder / decoder standard which surpasses VP8. This implies further future hardware changes, incompatibility, inability to playback video, forced upgrades, money and time wasted (as it will be the case when switching to VP8).

Using a different h264 profile would imply hardware updates as well in the case of mobile devices.

[rawr_boy81]

  On 22/05/2010 at 13:42, Saul Goodman said:

After reading the WebM FAQ, looks like the spec is final. Or am I misunderstanding something here?

You're missing so much you might as well give up whilst you're ahead. Tweaking and optimising does not mean being incompatible with the specification - you can tweak and optimise the algorithms to analyse each frame so that they're more efficient thus improve quality whilst maintaining the same amount of bandwidth.

[George P Global Moderator]

  On 23/05/2010 at 03:17, rawr_boy81 said:

You're missing so much you might as well give up whilst you're ahead. Tweaking and optimising does not mean being incompatible with the specification - you can tweak and optimise the algorithms to analyse each frame so that they're more efficient thus improve quality whilst maintaining the same amount of bandwidth.

Sure you can tweak it and gain on the performance side, that's all well and good but you'll run into a wall at some point. Then we'll all be looking at whatever the next new format will be. This has always been the case. Also with the move to more and more HD and faster connections for internet users the whole "base profile" or SD/low-res comparison is starting to get pointless. If you drop the bitrate and res low enough it doesn't really matter how magical your codec is the outcome will look like crap. I'm sure some other person can redo these tests using other tools and have h264 looking even better probably. I'd much rather have a HD comparison since that's where video is naturally headed, even on the web.

[torrentthief]

I'm hoping google works with nvidia and ATI and tries to get gpu encoding working for atleast some parts of the code so that the cpu and gpu can work together to encode. This would mean youtube videos would finish processing quicker after you have uploaded. It says on their website: "Look at GPGPU - offload encoding and decoding to the GPU where possible." So hopefully with the help of nvidia, ati, intel and via they might be able to pull it off.

I wonder if google will buy out CoreAVC and DiAVC, i'm sure they could be useful, maybe even open source them and release them for free. More and more people would be able to play HD videos on slower and older computers then.

[Boz]

  On 23/05/2010 at 08:03, GP007 said:

Sure you can tweak it and gain on the performance side, that's all well and good but you'll run into a wall at some point. Then we'll all be looking at whatever the next new format will be. This has always been the case. Also with the move to more and more HD and faster connections for internet users the whole "base profile" or SD/low-res comparison is starting to get pointless. If you drop the bitrate and res low enough it doesn't really matter how magical your codec is the outcome will look like crap. I'm sure some other person can redo these tests using other tools and have h264 looking even better probably. I'd much rather have a HD comparison since that's where video is naturally headed, even on the web.

It is absolute nonsense that "the person testing probably used a not so great codec for h.264".. What you are saying doesn't make sense.

First of all. .VP8 codec is brand new.. there are no real refined encoders. It's basic. While h.264 is already mature. It has nowhere to go in quality. The only thing here is the encoders and they are all for the most part mature.

I can almost guarantee you that once VP8 gets into the hand of open source people and give it a year or two you will have stunning results that will be better than h.264 (at least Baseline and Constrained Baseline profiles) and it will be close to Main/High.

[Richard Hammond]

Ive read of your predictions..... do you really think that h264 is going to stand still while VP8 is being developed?

[hdood]

  On 23/05/2010 at 00:55, Boz said:

No it doesn't' suffer from patent and license problems.. it's OPEN SOURCE. You are maybe hoping it will based on threats from MPEG-LA.. but it's purely speculative whether or not MPEG-LA actually even has a case here and at this point mostly fear mongering by those with vested interest in it with hopes to try to stop VP8.

I wish you would stop using terms that confuse people. There is only one difference between H.264 and VP8, and that is that Google offers you a perpetual royalty-free license to use the Google patents that cover it and trademarks owned by Google. That's the only difference, and that's what the advantage is.

It's not about "open source," although it is obviously useful to have a free open source implementation of it (which Google provides), especially since VP8 appears to be very messy and poorly documented, which makes it difficult for others to implement decoders. There are open source H.264 decoders and encoders as well. The issue is not one of source code, it's purely about patent and trademark licensing.

Being open source means nothing when it comes to patents. If I write a piece of software and give the source code away, that doesn't mean it isn't infringing on patents. The patents the MPEG-LA members have thrown into the patent pool are also not the only patents out there, and are not the only patents that might apply to VP8. Am I saying that it's certain that they'll be sued by someone? No. I'm just saying that you are oversimplifying it.

[torrentthief]

I'm sure google would counter-sue btw as i'm sure h264 violates some of the VP8 patents. I just hope they sign a document to say that they will or won't sue quickly and stick to that forever that way people can adopt the technology without fear of having to pay licences in the future.

[hdood]

  On 23/05/2010 at 09:33, torrentthief said:

i'm sure h264 violates some of the VP8 patents.

Why are you sure of that? Source?

[HawkMan]

  On 23/05/2010 at 08:51, Boz said:

It is absolute nonsense that "the person testing probably used a not so great codec for h.264".. What you are saying doesn't make sense.

First of all. .VP8 codec is brand new.. there are no real refined encoders. It's basic. While h.264 is already mature. It has nowhere to go in quality. The only thing here is the encoders and they are all for the most part mature.

I can almost guarantee you that once VP8 gets into the hand of open source people and give it a year or two you will have stunning results that will be better than h.264 (at least Baseline and Constrained Baseline profiles) and it will be close to Main/High.

Seriously Boz, Open Source has NOTHING to do with the quality of the codec, decoder or it's license ability. it can be open source and still require you to pay to MPEG-LA. Look at the code, parts of the code does pretty much the EXACT same thing as h264, down to the same size blocks and sub blocks. There some pretty obvious glaring infringements there.

but even so if we ignore that, it's not better than baseline. it's ALMOST as good as baseline, not as good, not better, and that was using a crap h264 encoder and not the good ones. the problem of course is motion, yeah, it's got a few other issues to, but mainly, it's motion, artifacting on motion and especially high speed motion. that's not good for a movie format.

And back to the open source thing, As I sad it doesn't matter if it's opn source, the spec itself is locked down, so it doesn't matter if the spec is open source, it's after all locked and can't be changed anyway. which makes it about as open as h264 anyway. the spec for that is also available for anyone to develop for. with the difference that h264 actually has a proper spec, not a paper that is merely copy pasted code snippets from the encoders c code, bugs and all. So Open source already has h264, x264 has been here for years, they've already been workign on that for years.

and as I explained to you before, open source doesn't magically mean better code. too many chefs, the core group won't accept submissions, and single persons won't be able to do much with it anyway. And again, do your seriously expect free hobby coders to make something of the same quality or better than the guys at paces like fraunhofer ?

besides codec quality, license issues and all that also isn't the only problem with accepting VP8 as a new webstandard. The main issue still remains, other devices, set top boxes, tv's, cell phones and so on, that won't be able to support this new format. It's ridiculous to have two formats when we have one that's already better and so ingrained in everything ith hardware decoding in practically everything today. But I guess you think people should just encode their videos twice, once to put on their homepage with the video tag and once to use on all their mpeg4 equipment.

Seriously, the whole idea of was to make things easier, that's not easier.

[torrentthief]

  On 23/05/2010 at 09:37, hdood said:

Why are you sure of that? Source?

because VP8 was developed using VP6 as it's original source and VP3 as the source before that, comments in the VP8 sourcecode date back to 2004 which is before h264 was created.

[hdood]

  On 23/05/2010 at 09:39, torrentthief said:

2004 which is before h264 was created.

Nope.

[boogerjones]

  On 23/05/2010 at 01:09, Luis Mazza said:

We don't need another video format if it's not better than H.264 with the same bitrate or with a lower bitrate and same quality just because of patents. All Windows, MAC OS an others have free options to decode H.264, so why bother with another format because it is patent-free? TODAY this is useless... Google: you should have done that back in 2004, at least.

Yo, what's up? How's the weather down there? And guess what? You have no idea what you're talking about. A patent-free codec with good hardware support would DOMINATE the web video market, even if it was of lower quality (though I think the arguments about quality in VP8 versus h.264 thus far are premature and total bull****). You need a microscope to notice any differences with fluid video when comparing these codecs. B-frames, PSYRD, trellis, whatever; nobody but lonely grad students developing video encoders in their free time gives a ****.

VP8 as good as h.264 baseline? Well, great, because that's the level that everything online gets encoded at anyway since Apple's products won't play anything with b-frames, CABAC, or trellis in it. The visual differences aren't noticeable and the compression benefit of the higher h.264 levels is pretty minimal for most video. Blu-ray is a different story, but how many of you are encoding commercial Blu-ray discs? Oh, right, probably none of you.

Should Google (or anybody else) have done this back in 2004? Yeah, that's exactly what the h.264 standards committee was doing. And they were ****ing smart, because look at the ********ery they've enabled by bringing 200 patent claimants together and scaring everyone away from developing a competing codec.

MPEG-LA won't go too hard after Google; that's not a winning strategy. The winning strategy is to strong-arm the hardware manufacturers to keep VP8 out of mobile devices. That market is HUGE and is expanding exponentially; it's where the big money is, kids. No sane hardware CEO is going to stand up at the annual investors meeting and say "Hey, let's risk everything and fight the MPEG-LA so we can keep this new, untested codec in our devices."

[Mathiasdm]

  On 22/05/2010 at 12:10, brentaal said:

First Look: H.264 and VP8 Compared

Hm. I'd like to see PSNR statistics and tests on high quality images (720p, 1080p) as well.

  On 22/05/2010 at 12:29, torrentthief said:

h264 will of course be better, but give it 12 months and i'm sure they will be equal by then now that it has been open sourced and dozen's of tech companies are involved with it. Lets hope someone makes a plugin for firefox to use windows 7 for h264 playback that way we can have videos in the same great quality as flash but with less cpu usage and crashes.

The specifications have been finalised.

  On 22/05/2010 at 13:00, ipodman715 said:

Great technical comparison

That one is very interesting. Too bad VP8 doesn't have B-frames. That'll matter a lot for the quality.

Also, this shows some of the problems with software patents. So many ways to improve the quality have been patented, it's a real shame.

  On 23/05/2010 at 08:51, Boz said:

It is absolute nonsense that "the person testing probably used a not so great codec for h.264".. What you are saying doesn't make sense.

First of all. .VP8 codec is brand new.. there are no real refined encoders. It's basic. While h.264 is already mature. It has nowhere to go in quality. The only thing here is the encoders and they are all for the most part mature.

I can almost guarantee you that once VP8 gets into the hand of open source people and give it a year or two you will have stunning results that will be better than h.264 (at least Baseline and Constrained Baseline profiles) and it will be close to Main/High.

If the specifications don't have some important items for quality (B-frames, for example), I doubt it'll get higher quality than h.264 or be close to the High profile.

Still, this codec can be useful (no royalties), but I fear for patent issues.