The H.264/AVC/MPEG4 discussion thread

[Eeknay]

The problem with the whole PS3 thing is that it's incredibly picky about even main profile stuff, and though they might have added high profile for m2ts the tools on the PC side author the stream incorrectly so it doesn't play. The sad thing is, if I take one of my own Sky backups, remux to MP4 and re-encode the audio to LC-AAC, the PS3 will; recognize the file, show the correct length, resolution, audio, filesize etc........ but WON'T play it. The sense, as they say, is unmaking. It's all very well they say the support these things, but if they don't flick the switch...

If you take a AVC m2ts stream decrypted from a Blu-Ray and bung it on there, it'll play. But homebrew authoring barely works for MPEG2, let alone VC1 or AVC which doesn't really work at all.

Also AFAIK 4.1 even at High Profile doesn't have certain options that most subbers use (so far I see p4x4? I'm making a big assumption someone would just use "All" for those macroblock options, or some might even use those megui profiles which I think just use Unrestricted? Maybe? Correct me if I'm wrong) so they'd need re-encoding anyway. At least what I encode (and have encoded for Arienai in the past... >_>) doesn't conform to 4.1.

I'm grabbing at straws here, but I would still hope that ASP is supported despite what they say. Supporting SP and not ASP, especially when you have H.264 support is dumb

I muxed a 960x540 XviD (scene release so no funky options methinks) with re-encoded LC-AAC to MP4 and tried it on the PS3 recently; no go. That's just one file though, I'm gonna try some more I think.

Really though, how hard can it be to quench peoples thirst when XBMC has been doing it for several years sans [decent] h264 support?

[Quarkboy]

Originally Posted by Eeknay

Also AFAIK 4.1 even at High Profile doesn't have certain options that most subbers use (so far I see p4x4? I'm making a big assumption someone would just use "All" for those macroblock options, or some might even use those megui profiles which I think just use Unrestricted? Maybe? Correct me if I'm wrong) so they'd need re-encoding anyway. At least what I encode (and have encoded for Arienai in the past... >_>) doesn't conform to 4.1.

I think you're getting confused... 4.1 is the LEVEL, and high is the profile.
The level is basically a throughput limit, it limits the resolution, max bitrate, max framerate, etc. It has nothing to do with the p4x4 transform at all.
4.1 has a max resolution of 2048x1024 at 30 fps, which is plenty for any fansub. Actually, I think that my 1280x720 releases are only Level 3.1, although I don't explicitly set that limiting option, I highly doubt with the low bitrates used it crosses any of the limits.


High profile is defined to support p4x4 and basically all the options that fansubbers use in x264, see Zero's chart on the previous page.
The only things not supported in high profile are options like flexible macroblock ordering and SP and SI slices, which x264 doesn't even implement.


So unless Sony simply lies when they say it supports true high profile, or they've purposefully disabled certain things.

What you've encoded for me most certainly DOES conform to High Profile Level 4.1, as has almost everything ever encoded for a fansub.

[Eeknay]

Originally Posted by Quarkboy

I think you're getting confused...

Yeah, basically. Ignore me

What you've encoded for me most certainly DOES conform to High Profile Level 4.1, as has almost everything ever encoded for a fansub

Ok. So it's a matter of container/audio and whether or not they flick the switch. So technically if the ES from a h264 fansub was remuxed somehow to m2ts, it *technically* should work according to Sony? I'm thinking of mp42ts and TSRemux tag team, but I just read that mp42ts doesn't do AVC. Hrn.

[Quarkboy]

Originally Posted by Eeknay

Ok. So it's a matter of container/audio and whether or not they flick the switch. So technically if the ES from a h264 fansub was remuxed somehow to m2ts, it *technically* should work according to Sony? I'm thinking of mp42ts and TSRemux tag team, but I just read that mp42ts doesn't do AVC. Hrn.

Yup. "If you mux it, it will play" as James Earl Jones once said...

Frankly, I think it'd be easier to wait for them to simply support it in mp4 then figure out how to mux m2ts files with software that barely exists.

One thing I'm not too sure about when it comes to level... Is there somewhere in the stream itself that contains the level data? Or is it more of a, "If it conforms the hardware should have no problems playing it, otherwise you're on your own" type thing. I would hate for my encodes not to be played simply because a bit that says "Level XX" wasn't set, even if it does conform to the standards.

[Eeknay]

Originally Posted by Quarkboy

Frankly, I think it'd be easier to wait for them to simply support it in mp4 then figure out how to mux m2ts files with software that barely exists.

True, but it's fun fooling around. My endgame isn't so much fansubs as it is HD caps (even then it wouldn't be a permenant solution, just a one off for certain shows or movies now and again). In the end I have to remux no matter what from a TS to something else.

One thing I'm not too sure about when it comes to level... Is there somewhere in the stream itself that contains the level data? Or is it more of a, "If it conforms the hardware should have no problems playing it, otherwise you're on your own" type thing. I would hate for my encodes not to be played simply because a bit that says "Level XX" wasn't set, even if it does conform to the standards.

Good point.

[Zero1]

Originally Posted by Eeknay

The sad thing is, if I take one of my own Sky backups, remux to MP4 and re-encode the audio to LC-AAC, the PS3 will; recognize the file, show the correct length, resolution, audio, filesize etc........ but WON'T play it.

That's pretty damn poor. Why is it such a task to write a spec compliant decoder?

Originally Posted by Eeknay

Also AFAIK 4.1 even at High Profile doesn't have certain options that most subbers use (so far I see p4x4? I'm making a big assumption someone would just use "All" for those macroblock options, or some might even use those megui profiles which I think just use Unrestricted? Maybe? Correct me if I'm wrong) so they'd need re-encoding anyway. At least what I encode (and have encoded for Arienai in the past... >_>) doesn't conform to 4.1.

Well as Quarkboy later went on to explain, the profiles define the encoding tools, and levels the complexity of the video in terms of resolution, frame rate, bitrate and decoded picture buffer (DPB).

I find that the majority of SD H.264 falls within level 3.1 or 3.2; less complex encodes may get away with level 3, but on the other hand so called extreme settings in x264 can leave you with level 4 SD H.264. One of the main factors I find is the number of reference frames which pushes up the required decoded picture buffer. I haven't done much HD encoding, mainly because good source is hard to come by (forget upscales, damn waste of time); but that's where it gets interesting.

Originally Posted by Eeknay

I muxed a 960x540 XviD (scene release so no funky options methinks) with re-encoded LC-AAC to MP4 and tried it on the PS3 recently; no go. That's just one file though, I'm gonna try some more I think.

I'm not making excuses (especially not for people who make whack decoders, see Apple & Sony) but was it ISO ASP, or was it the VfW bastardisation remuxed? It might not even matter in the end since MP4Box probably unbastardises it, but I'm just making certain. Also while it may or may not have any bearing, it's also a non mod16 resolution.

Originally Posted by Quarkboy

So unless Sony simply lies when they say it supports true high profile, or they've purposefully disabled certain things.

It wouldn't be the first time. The PSP used to have such constraints applied for resolution and framerates. Although UMD movies were in the region of 480x272 (IIRC), user encodes were only allowed to be 320x240 or something in that region.

Originally Posted by Quarkboy

One thing I'm not too sure about when it comes to level... Is there somewhere in the stream itself that contains the level data? Or is it more of a, "If it conforms the hardware should have no problems playing it, otherwise you're on your own" type thing. I would hate for my encodes not to be played simply because a bit that says "Level XX" wasn't set, even if it does conform to the standards.

Yes, the profile/level is stored in the bitstream like it was for ASP, and even MPEG-2 (IIRC). You may have spotted it before, but there is an option within x264 to specify the level of your encode, and if you set it too low, it will give you a warning, although this isn't gospel; I don't think it takes into account things such as VBV but I may be wrong.

There may even be a tool for messing with bitstream attributes, I'm pretty sure Vqual/Tektronix will have something, but that's mega expensive software aimed at companies developing H.264 codecs. Failing that, it should be quite easy to hex edit the bitstream to change the profile/level to how you want, but I haven't looked that far into it yet, I've only just got my head around basic MPEG-2 attribute editing.

I have a feeling that most, if not all hardware or software will pretty much ignore the level in terms of refusing to play it or not, I suspect it will try to play it and either do so, or struggle.

[Quarkboy]

Originally Posted by Zero1

I have a feeling that most, if not all hardware or software will pretty much ignore the level in terms of refusing to play it or not, I suspect it will try to play it and either do so, or struggle.

Here's hoping .

I believe that both Blu-ray and HD-DVD players should be able to play h.264 high profile @L4.1, correct? Let's hope that these super-cheap (~$300) next-gen DVD players they are starting to manufacture in China that I've heard rumors of aren't as picky about container formats as the name brands are.

[Nicholi]

Originally Posted by Eeknay

The sense, as they say, is unmaking.

Sorry for the random off topic'ness... but I simply must know where else you heard that saying n_n!

As I recall MP4Box should by default remove packed bitstream and such nonsense. Making an MP4 with VfW content still inside requires some effort :P.

[Eeknay]

ahoy darkhold kkkrew?

I'm not making excuses (especially not for people who make whack decoders, see Apple & Sony) but was it ISO ASP, or was it the VfW bastardisation remuxed? It might not even matter in the end since MP4Box probably unbastardises it, but I'm just making certain. Also while it may or may not have any bearing, it's also a non mod16 resolution.

Probably VfW. I'll try encoding a native ASP sample (mod16 too this time) now and see what happens, although I doubt it'll work...

EDIT: Tried four samples; 1088p, 1080p, 720p, 400p, none of them work. It recognizes the length but no other info (just like the scene VfW test).

[Zero1]

GJ Sony, way to not support one of the most common codecs ever.

[ender]

Just be happy they didn't try to push a video ATRAC

[YourLocalGP]

Zero1, thatnks for your terrific help! Unfortauntely something's not happening correctly with the files so the output file never appears. After "Extracting MPEG-4 AAC", the "file1.txt" appears with the following info in it:

"09/05/2007
15:46
* Movie Info *
Timescale 600 - Duration 00:13:55.880
Fragmented File no - 2 track(s)
File Brand mp42 - version 1
Created: GMT Fri Apr 13 14:53:36 2007

File has no MPEG4 IOD/OD

Track # 1 Info - TrackID 1 - TimeScale 44100 - Duration 00:13:55.941
Media Info: Language "English" - Type "soun:mp4a" - 36001 samples
MPEG-4 Config: Audio Stream - ObjectTypeIndication 0x40
MPEG-4 Audio AAC LC - 1 Channel(s) - SampleRate 44100
Self-synchronized

Track # 2 Info - TrackID 2 - TimeScale 600 - Duration 00:13:55.880
Media Info: Language "English" - Type "vide:avc1" - 20897 samples
MPEG-4 Config: Visual Stream - ObjectTypeIndication 0x21
AVC/H264 Video - Visual Size 320 x 180 - Profile Main @ Level 1.2
NAL Unit length bits: 32
Synchronized on stream 1"

Thereafter, the batch file window gives me:

"file h264parse.exe not found
Could Not Find C:\Temp\mp4_research\h264parse\file1_track1.h264
Extracting MPEG-4 AAC
file h264parse.exe not found
Could Not Find C:\Temp\mp4_research\h264parse\file2_track1.h264

7-Zip 4.42 Copyright (c) 1999-2006 Igor Pavlov 2006-05-14


Error:
There are no installed archive handlers"

H264parse.exe is definitely there; I can run it manually from the command line. I think it must mean that the file it is looking for is not found....

[Zero1]

Originally Posted by ender

Just be happy they didn't try to push a video ATRAC

Actually the PSP UMD movie discs use H.264 video with ATRAC3 audio... Seems they don't know when to quit.

Originally Posted by YourLocalGP

Zero1, thatnks for your terrific help! Unfortauntely something's not happening correctly with the files so the output file never appears. After "Extracting MPEG-4 AAC", the "file1.txt" appears with the following info in it:

We may be on to something here. Firstly the reason the batch likely failed is because I told it to extract track 1 in the MP4, which is usually the video. For whatever reason the file that gave the input you posted has the video as track 2, and the audio as track 1, which means h264parse.exe is trying to parse AAC audio which fails, and then 7zip has nothing to archive; it's like a chain reaction failure.

I don't know if you are savvy with batch files, but all you need to do open the batch in a text editor like notepad, and look for the line:

mp4box.exe -raw 1 "file1.mp4"

Simply change -raw 1 to -raw 2; all this does is specify which TrackID to extract. You can check the TrackIDs beforehand manually with:

mp4box.exe -info "file1.mp4"

Also from this little bit of information MP4Box has given me, we might have found the cause. Pay attention to the line:

AVC/H264 Video - Visual Size 320 x 180 - Profile Main @ Level 1.2

Now if this is for an iPod; iPod's supposedly support baseline profile @ level 1.3. The fact that this video is level 1.2 is irrelevant (it just means it will use less CPU), but notice how MP4Box is reporting the profile as Main. This may be the issue. For the encode to be main profile, it's likely it's using B-frames or CABAC. For an encode to be baseline, you need to disable B-frames and weighted prediction (wpred is a B-frames option, so disabling B-frames disables wpred too), and use CAVLC instead of CABAC for the entropy coding. Most encoding applications should give you the choice of changing these options, x264 certainly does.

Well try encoding it with those suggestions and see if you have any joy. If it does still not work, make the suggested changes to the batch file and see if you can get that output for me again. I think we may have it this time though.

Also it's perhaps of little consequence as to whether the file works or not; but I'd suggest cropping the video to 320x176 so it's mod16. Some decoders are known to mess up with odd size resolutions, so I'm just making sure we eliminate as many things as possible.

[martino]

I wonder one thing though, even if Sony doesn't support playback of MKV video files, with ASS softsubs, or just some level/profile of h264, there is a workaround (I presume). Since the PS3 allows an OS, Linux in this case, to be installed and booted into, there isn't that much stopping the user from opening that file in mplayer. However, here I don't know how much that Linux distribution (Yellow Dog as official, though FC worked too IIRC) can do when it comes to interfacing with display and sound devices. In addition to that it's not like mplayer is exactly the most user friendly player, more like the opposite (who doesn't like nice clickable buttons...). Then there's the issue of the hardware architecture, in this case mostly the CPU. Can the application take advantage of all those processing units? Is it optimised for it? Not like that would be such an awful problem, you just need to make a few people aware of it and you never know what's around the corner. After all, wasn't x264 already optimised for it to some extent, or least there was an attempt to...

Though I guess this wouldn't really apply to the 360, unless someone already managed to boot Linux on it. I stopped keeping track of all this console stuff lately.

Hopefully my post makes some sense...
<--- Not fully sober at the time of writing this post

[jfs]

Now, there are Linux-applications that provide sensible user interfaces for PVR-like applications, it should be possible to get some of those to work. At least if it's possible to get any kind of (reasonably fast) framebuffer graphics mode. Or maybe port XBMC?

[Eeknay]

I did try installing Linux for that purpose (testing playback) but it didn't recognize DSL (Damn Small Linux) or the latest Mandrake. I don't know if there's only a select few it likes or if I suck or what.

[emptyeighty]

Originally Posted by Eeknay

I did try installing Linux for that purpose (testing playback) but it didn't recognize DSL (Damn Small Linux) or the latest Mandrake. I don't know if there's only a select few it likes or if I suck or what.

Try one of those listed here.

[xat]

Originally Posted by martino

However, here I don't know how much that Linux distribution (Yellow Dog as official, though FC worked too IIRC) can do when it comes to interfacing with display and sound devices.

Sound seems to be working fine on this end -- haven't tried it with more than 2 channel audio though since I don't have a setup for it. There's more in terms of choices too. I'm seeing Gentoo and Ubuntu being used more and more, and I use Gentoo myself. That and the lightweight XFCE on top of it amount to a relatively small footprint.

In addition to that it's not like mplayer is exactly the most user friendly player, more like the opposite (who doesn't like nice clickable buttons...).

Mplayer does have a gui supplement though. Failing that, there's also VLC which works okay as well.

Then there's the issue of the hardware architecture, in this case mostly the CPU. Can the application take advantage of all those processing units? Is it optimised for it? Not like that would be such an awful problem, you just need to make a few people aware of it and you never know what's around the corner.

Yes, no. The cell is pretty much the only critical hardware that isn't virtualized, but things like ffmpeg have yet to be optimized for it (as far as I know). The ps3/cell could handle HD h264 content easily, but not through linux at present. This may have changed as of Ubuntu 7.04, but I haven't been able to test that yet.

After all, wasn't x264 already optimised for it to some extent, or least there was an attempt to...

Sorta. Performance is lackluster for now. Hopefully it's only a matter of time before we see (vast) improvements in performance. I really want to use my ps3 for encoding.

[Nicholi]

Start a "buy a PS3 for pengvado" fund n_n. I'm sure he might be more willing to optimize it if he can easily test it himself :P.

[Zero1]

I don't know about you guys, but I'd prefer to see more quality tweaks and x86 optimisations before wasting time with PS3.

PS3's hardware will never change, so you are stuck with a set CPU speed; however with a nice optimised x86 build, CPUs are always getting faster and you will feel the improvements for years to come.

I also doubt you will get the full potential of the PS3 too, since Sony seem to like locking down their hardware.