VP8, VP9 WebM encoding feature. Theora

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

*******************************
DaveFilmsUS
Posts: 1
Joined: Tue Jan 07, 2014 12:50 am

VP8, VP9 WebM encoding feature. Theora

Post by DaveFilmsUS »

Please add the WebM, VP8 / VP9 codec with MKV container. It would be very helpful to support Theora too. Since it is a free codec..
Pls add the option in next release. :D

User avatar
Rodeo
HandBrake Team
Posts: 12534
Joined: Tue Mar 03, 2009 8:55 pm

Re: VP8, VP9 WebM encoding feature. Theora

Post by Rodeo »

We already support Theora.

Next version will most likely have VP8 too (we may decide to remove Theora once we decide VP8 support is usable enough).

VP9 will have to wait for muxer support, and was so slow last time we checked, it's not usable anyway.

Ayoross
Posts: 48
Joined: Mon Jun 11, 2012 9:07 pm

Re: VP8, VP9 WebM encoding feature. Theora

Post by Ayoross »

How slow ?

mduell
Veteran User
Posts: 7151
Joined: Sat Apr 21, 2007 8:54 pm

Re: VP8, VP9 WebM encoding feature. Theora

Post by mduell »

Minutes per frame. Days per minute.

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

Re: VP8, VP9 WebM encoding feature. Theora

Post by TedJ »

Three months to encode a full feature film... that slow.

SkOrPn
Posts: 1
Joined: Fri Jan 31, 2014 7:27 pm

Re: VP8, VP9 WebM encoding feature. Theora

Post by SkOrPn »

Hmm, VP9 is supposedly 7% faster than H.265, or so I read somewhere. YouTube is actively converting most of their HD to it as we speak, starting with 4k videos, then 1080p, and then 720p, and then finally SD content. However, they do not have super computers, in fact, they have nothing more powerful standard high grade work stations, probably with i7's. Surprised they would bother to convert video if it takes them weeks or months to do a single video. Not to mention Netflix is actively converting their content to HEVC which is supposedly slower than VP9. However, I have no clue if this is true or not, as Its just what I read from info that was gathered at CES.

Samsung, Sony, LG and Vizio all said they will have HEVC, VP9 capable hardware being sold by March-April 2014. I am hopeful that the Samsung models will have the Plex player, so we can use the built in HEVC or VP9 hardware decoding for content we have stored onto our Media Servers. That is why I just built a Intel Hexa Core XEON based Media Server to prepare for HEVC decoding, lol. Not that I would use it right away, but the wife wants the DVD's either ripped and stored onto hard drives or sold/given away in a garage sale, and we have thousands I do not want to lose. All our BD disks are already ripped and taking up 10gb per 2 hours of video. Sounds like I have a few hundred years of work a head of me for the DVD's though. lol

Oh, Vizio said effective immediately they are no longer making 3D TV's as they have solely committed all R&D towards 4K and the decoding of high-efficiency codecs on their TV sets. Even Netflix said they are committing all budget on the concentration of 4K and HEVC streaming content. However, its not as important for us home users since most of the time we will be streaming our personal content over our own networks, and not over the Internet. If you do decide to stream content over the Internet HEVC or VP9 would be a HUGE help. I'm just interested in the hard drive space savings myself. Would love to have BD movies down to 5-6gb and DVD's around 2gb with no less PQ loss. I hope I see that day very soon...

User avatar
Rodeo
HandBrake Team
Posts: 12534
Joined: Tue Mar 03, 2009 8:55 pm

Re: VP8, VP9 WebM encoding feature. Theora

Post by Rodeo »

SkOrPn wrote:Hmm, VP9 is supposedly 7% faster than H.265, or so I read somewhere.
Is diesel faster than gas? One encoder can be faster than another - same goes for a decoder - video coding standards, on the other hand, are just specifications, and have no speed.
SkOrPn wrote:YouTube is actively converting most of their HD to it as we speak, starting with 4k videos, then 1080p, and then 720p, and then finally SD content. However, they do not have super computers, in fact, they have nothing more powerful standard high grade work stations, probably with i7's. Surprised they would bother to convert video if it takes them weeks or months to do a single video.
They probably have a [Censored] of them though.
SkOrPn wrote:I'm just interested in the hard drive space savings myself. Would love to have BD movies down to 5-6gb and DVD's around 2gb with no less PQ loss. I hope I see that day very soon...
Better standards are one thing… better encoders are another. We'll probably have to wait quite a while longer for HEVC and VP9 encoders to mature enough to beat x264 consistently.

User avatar
s55
HandBrake Team
Posts: 9786
Joined: Sun Dec 24, 2006 1:05 pm

Re: VP8, VP9 WebM encoding feature. Theora

Post by s55 »

Just to extend on what Rodeo is saying, There are h.264 encoders other than x264 and some of them are 2x or more slower doing the same task.

x265 has an active development team working on it that are racing forward at the moment, so it seems to have overtaken VP9 in terms of speed. I think john almost got real-time h.265 output from it from a very beefy box. So while you could still be talking a day + on an average machine on faster settings, it's far closer to being usable at the moment than VP9.

That said, VP9 is also constantly improving, but it seems to be at a slower pace in terms of development. Even VP8 is still very slow compared to x264 and in theory, you could make it just as fast.

mduell
Veteran User
Posts: 7151
Joined: Sat Apr 21, 2007 8:54 pm

Re: VP8, VP9 WebM encoding feature. Theora

Post by mduell »

SkOrPn wrote:Hmm, VP9 is supposedly 7% faster than H.265, or so I read somewhere. YouTube is actively converting most of their HD to it as we speak, starting with 4k videos, then 1080p, and then 720p, and then finally SD content. However, they do not have super computers, in fact, they have nothing more powerful standard high grade work stations, probably with i7's.
I'll go out on a limb and guess they're using servers with Xeons, rather than some crap i7 workstation (no ECC, lol).

User avatar
JohnAStebbins
HandBrake Team
Posts: 5581
Joined: Sat Feb 09, 2008 7:21 pm

Re: VP8, VP9 WebM encoding feature. Theora

Post by JohnAStebbins »

SkOrPn wrote:Hmm, VP9 is supposedly 7% faster than H.265, or so I read somewhere. YouTube is actively converting most of their HD to it as we speak, starting with 4k videos, then 1080p, and then 720p, and then finally SD content. However, they do not have super computers, in fact, they have nothing more powerful standard high grade work stations, probably with i7's. Surprised they would bother to convert video if it takes them weeks or months to do a single video. Not to mention Netflix is actively converting their content to HEVC which is supposedly slower than VP9. However, I have no clue if this is true or not, as Its just what I read from info that was gathered at CES.
Lets talk specific encoders. I've tested both x265 and libvpx. x265 on a 6 core sandy bridge was getting about 5 frames per second. libvpx was getting about 1 frame per minute. A couple orders of magnitude difference in speed with x265 being faster.

Regarding googles conversion to vp9, I wouldn't be at all surprised if they had an internal home grown encoder that is faster than libvpx. They are a business after all and a faster encoder would be a competitive advantage. And as you have pointed out, they have lots and lots of fast hardware.

I don't know where you came up the that 7% VP9 speed advantage statistic, but it sounds like something someone threw out there with no supporting evidence and then it got propagated by all the freetards of the world. It is certainly possible that an implementation of VP9 could be faster than an implementation of x265. But speed is mostly implementation specific, and no such implementations are available to us mere mortals. A theoretical 7% speed advantage is trivial to overcome with a better implementation.

User avatar
Rodeo
HandBrake Team
Posts: 12534
Joined: Tue Mar 03, 2009 8:55 pm

Re: VP8, VP9 WebM encoding feature. Theora

Post by Rodeo »

Probably nothing in software, Google would probably have found a way to port it to libvpx… it's the same company. But I hear they're really pushing hardware vendors to support VP9 in hardware, it's not impossible they'd have access to some yet-unreleased VP9-dedicated silicon.

This is all just guesswork though.

John, were your latest tests with 1.3.0+? Apparently libvpx speed has gotten "reasonable" lately (on super-beefy hardware, that is).

User avatar
JohnAStebbins
HandBrake Team
Posts: 5581
Joined: Sat Feb 09, 2008 7:21 pm

Re: VP8, VP9 WebM encoding feature. Theora

Post by JohnAStebbins »

I don't recall what version it was. But the release was accompanied by some grandiose claims of vast speed improvements (i.e. 10x). My testing using the same test script I used on previous versions showed almost zero speed improvement.

K@mpfH@mster
Posts: 36
Joined: Tue Jul 15, 2014 10:42 am

Re: VP8, VP9 WebM encoding feature. Theora

Post by K@mpfH@mster »

Hi, could you remove Theora in nightlys, it is an outdated and nearly unsupported video codec. VP8 is much better for web

VP9
Posts: 20
Joined: Sat May 03, 2014 12:29 am

Re: VP8, VP9 WebM encoding feature. Theora

Post by VP9 »

K@mpfH@mster wrote:Hi, could you remove Theora in nightlys.
Why not leave Theora as an option?

It is OK that VP8 can provide more compression/quality but Theora is much faster to encode. VP8 is significantly (I'd say much) slower than x264.
K@mpfH@mster wrote:it is an outdated and nearly unsupported video codec. VP8 is much better for web
The codec is there for anyone to download and use (for free ($)).

If it is unsupported than you mean it is unsupported on proprietary operating systems like Windows and MAC and in websites that do not wish to use it. Just download the codec for free and play, it is as simple as that. Most linux based system support it by default, and it is like WMV is for Windows.

They do not come with them by default because they can't simply support free codecs for their own financial benefit.

User avatar
JohnAStebbins
HandBrake Team
Posts: 5581
Joined: Sat Feb 09, 2008 7:21 pm

Re: VP8, VP9 WebM encoding feature. Theora

Post by JohnAStebbins »

We *do* plan to remove theora. Haven't really discussed whether it will be in the upcoming release or not. But I'm all for ditching it now.

nexradix
Posts: 25
Joined: Sat Sep 26, 2009 1:55 am

Re: VP8, VP9 WebM encoding feature. Theora

Post by nexradix »

Please consider leaving Theora in. I use it exclusively with Handbrake, as it is supported out-of-the-box on any mainstream GNU/Linux distro.

VP8 is great, and hopefully will overtake Theora as the default video codec in most distros. But until it does, I'll be using Theora.

Handbrake is currently one of the few apps that will *properly* and consistently produce Theora/Vorbis in MKV without lip-sync or other issues. Please don't kill off that ability just yet.

Djfe
Bright Spark User
Posts: 178
Joined: Tue May 13, 2014 8:01 pm

Re: VP8, VP9 WebM encoding feature. Theora

Post by Djfe »

SkOrPn wrote:Hmm, VP9 is supposedly 7% faster than H.265, or so I read somewhere.
Well they claimed that for VP8 in comparison with H264 already. Which obviously didn't end up right... H264 - as implemented in x264 - is way faster and is way more efficient than VP8 in any implemented form.
We will see whether a VP9 encoder is able to beat x264.

But H265/HEVC in x265 will do that for sure.

About Theora: Does it hurt to leave it in? -> From the amount of space it needs probably not, but what about the effort that is needed to maintain it? Is there any need to update the GUI in the future?
If not then I don't think, that it is neccessary to leave it out.

But in the end it's totally up to the dev.

If there is any need for Theora, then the people that want to use it can still use FFMPEG, an older handbrake version or another encoder

@DaveFilmUS you need to use MKV for Theora, MP4 doesn't support it.

User avatar
JohnAStebbins
HandBrake Team
Posts: 5581
Joined: Sat Feb 09, 2008 7:21 pm

Re: VP8, VP9 WebM encoding feature. Theora

Post by JohnAStebbins »

nexradix wrote:Please consider leaving Theora in. I use it exclusively with Handbrake, as it is supported out-of-the-box on any mainstream GNU/Linux distro.

VP8 is great, and hopefully will overtake Theora as the default video codec in most distros. But until it does, I'll be using Theora.

Handbrake is currently one of the few apps that will *properly* and consistently produce Theora/Vorbis in MKV without lip-sync or other issues. Please don't kill off that ability just yet.
We had a quick discussion on IRC about this. We will be leaving it in for the next release. But it will most likely be removed for the release after that. So nothing to worry about for at least another year :wink:

VP9
Posts: 20
Joined: Sat May 03, 2014 12:29 am

Re: VP8, VP9 WebM encoding feature. Theora

Post by VP9 »

JohnAStebbins wrote:So nothing to worry about for at least another year.
But MPEG4 and MPEG2 that is quite old and obsolete will be kept forever. Can you understand this? I do not (but I have theories I do not want to share).

User avatar
Rodeo
HandBrake Team
Posts: 12534
Joined: Tue Mar 03, 2009 8:55 pm

Re: VP8, VP9 WebM encoding feature. Theora

Post by Rodeo »

VP9 wrote:
JohnAStebbins wrote:So nothing to worry about for at least another year.
But MPEG4 and MPEG2 that is quite old and obsolete will be kept forever. Can you understand this?
Pretty simple really. Both encoders are really fast, and used by developers for testing purposes because of that (i.e. when wanting to test something where the video track is not relevant). Same can't be said for Theora, by a large margin.

Also, both MPEG encoders are part of libavcodec, which is essential for decoding and audio encoding, so getting rid of these encoders would bring zero benefit (we'd have to keep all the code anyway, since it's shared with VP8), whereas Theora would allow us to drop one library, simplify a bit of code, and reduce the maintenance workload (if only very slightly).

rrunner64
Posts: 2
Joined: Thu Sep 04, 2014 8:04 am

Re: VP8, VP9 WebM encoding feature. Theora

Post by rrunner64 »

Using HandBrake svn6388 with graphical user interface (64bit Ubuntu) the encoder uses only a single thread for VP8. Is this intended behaviour? I think multiple threads would increase the speed of encoding very much.
http://www.webmproject.org/docs/encoder ... and-decode

mduell
Veteran User
Posts: 7151
Joined: Sat Apr 21, 2007 8:54 pm

Re: VP8, VP9 WebM encoding feature. Theora

Post by mduell »

rrunner64 wrote:Using HandBrake svn6388 with graphical user interface (64bit Ubuntu) the encoder uses only a single thread for VP8. Is this intended behaviour? I think multiple threads would increase the speed of encoding very much.
Activity log?
Last edited by mduell on Fri Sep 05, 2014 3:37 am, edited 1 time in total.

moneymatt4life
Veteran User
Posts: 436
Joined: Fri Mar 09, 2012 5:26 am

Re: VP8, VP9 WebM encoding feature. Theora

Post by moneymatt4life »

i think it is intended, as i was getting low cpu usage with VP8 when i just tested it. so you have to specify the threads in the advanced options.

EDIT:
Rodeo: Nope, a bug. Should be fixed: https://trac.handbrake.fr/changeset/6389
JohnAStebbins:It was not intended. Just a bug nobody ever noticed because VP8 support is new in HandBrake and nobody really cares about it :wink:
ah... lol well count me in with the ones that don't really care about it :P
Last edited by moneymatt4life on Thu Sep 04, 2014 6:55 pm, edited 1 time in total.

User avatar
Rodeo
HandBrake Team
Posts: 12534
Joined: Tue Mar 03, 2009 8:55 pm

Re: VP8, VP9 WebM encoding feature. Theora

Post by Rodeo »

Nope, a bug. Should be fixed: https://trac.handbrake.fr/changeset/6389

User avatar
JohnAStebbins
HandBrake Team
Posts: 5581
Joined: Sat Feb 09, 2008 7:21 pm

Re: VP8, VP9 WebM encoding feature. Theora

Post by JohnAStebbins »

moneymatt4life wrote:i think it is intended, as i was getting low cpu usage with VP8 when i just tested it. so you have to specify the threads in the advanced options.
It was not intended. Just a bug nobody ever noticed because VP8 support is new in HandBrake and nobody really cares about it :wink:

Post Reply