Codecs: Libavcodec progress thread
 

New setting available:

- New adaptative bframe decision
Smaller size with higher quality at constant quant. Bframe decision is very better.

- New Bframe refine
Quality for bframe is improved.

+ 0.2 dB for OPSNR for my MPEG2 file reference
+ 1.0 Pt for SSIM for my MPEG2 file reference

yeah, but maybe libavcodec should first be 100% dvd compliant.

I made a 1 minute long clip using libavcodec, and verified the MPEG stream. It gave MANY error, some of them I quoted below:

This goes on for a while, totaling too much (different) errors. Bear in mind that I made a 1 minute testclip, which was 10MB.

Seems like you have a metric/objective test here which proves libavcodec has a long way to go :wink:

But remember: he is always right and there's no point in trying to make him believe he's not right... :?

First let it be 100% MPEG2 compliant. After we will see for DVD ;)

Which tool are you using ?

You can upload this 1 min source please:

- your source
- your avs script
- your bitrate
- your mencoder script
- tool for MPEG stream check

Oh please!! Cut us some slack...
Even if he tell's us what he used you will never stand corrected anyway :roll:

I want source just for reproduce MPEG stream error. I can't find hard source to produce no compliant stream for my different SAP with mencoder. This source and tool for check mpeg stream too are really interessing for me ...

:lol:

You're right :D

I'm using the Philips DVD Video Verifier tool. It can analyze MPEG2 files or
complete authored DVD's. For this test, I just analyzed the MPEG2 file.

Doesn't matter, just encode a clip with libavcodec :wink:

Hey, that is known as the holy graal of DVD verifying tools :D.
No way we can be wrong about libavcodec now, right? :lol:

You're absolutely right :D
I thought, let him see the metrics about it, which seems one of his favorite ways to test a video sample :lol:

Check this out, I just made the same clip with CCE, below are the full results 8)

Speaks for itself, where libavcodec has 19361 errors, CCE has none 8)
But, CCE has more syntax errors, this is one error that occurs 103 times.


EDIT: added results for TMPG and HCenc:

HCenc:

TMPG:


Interesting to see that all three encoders have the same "error 1106". In CCE the quantity of this error is the most.

BTW, a screenshot from bitrateviewer which also need no words.
More objective than this isn't possible :)

Encoded using a libavcodec, max bitrate at 8000kbps, avg bitrate at 5000kbps. As you all can see, the bitrate spikes are through the roof, and SAP's will crash.

http://www.digitalfaq.com/archives/error.gif

The bitrate peak was already known, but the results from this MPEG2 analyzer are simply amazing :D

1) Bitrate viewer can't help because maxbitrate is not frame bitrate but little part to complex vbv limitation. Bitrate can be ponctually very more than 9800 Kbps.

2) DVD-Video Verifier v1.5.0 is a very good tool for check DVD compliance. Now I can prove that Mencoder can make DVD MPEG2 Compliant stream:

I choose very hard source for Rate Control. IMO the hardest availables sources are 1080p Quick Time Trailer. I choose hard source with high motion and full real 16/9 screen like night watch trailer.

But if you want I can choose an other very hard source here ...
http://www.apple.com/trailers/

... or make complete encoding with DVD

Night watch trailer is very uncompressible source : your KDVD-720x576-Full-D1-PAL profil, CQ 70, Highest quality, Notch Matrix, Max Bitrate = 9500 Kbps done Video Elementary Stream at 3000 Kbps. I choose 3000 Kbps for make encoding withMencoder. I mux m2v stream with sonic scenarist (scenarist can check m2v DVD compliance too) or TMPG DVD Author to obtain vob+ifo and I check structure with DVD-Video Verifier v1.5.0:


I make the same check with TMPGEnc encoding:


Conclusion: My Mencoder Stream with my Mencoder Setting (Multipass, quantizer blur, quantizer range, highest quality setting, HVS optimisation ...) are perfectly MPEG2 DVD compliant without MPEG2 fatal error like underflow. Download my sample and try to find incompatible SAP if you want.


@ gamma

- thank's gamma for this link to DVD-Video Verifier
- Upload your little clip source if you want ... I will make compliant libavcodec encoding

@dailhot
I'm liking this tool too :D
I think it's very interesting to see that even the accepted/proven encoders produce errors. The ultimate test I think is still to burn on RW
and put it in the SAP.


@sagittaire

1) I find it very hard to believe that after all this mpeg2-compliancy discussions, you suddenly (within a day) bought the Philips DVD Verifier software, after I posted results here. I'm strongly thinking that you used w@rez for this.

2) OK, now burn it on a RW and put it in your SAP.

3) Nice results. I wonder why they differ so much from my findings. I find it hard to believe that you produced a stream with totally no errors. Verify a original (as in: bought) DVD, you'll see what I mean.

4) Have you made full-D1 or half-D1?

5) For the record: I posted no link, I don't want any affiliation with you obtaining this tool at probably an illeg@l way.

6) I must admit, encodings with libavcodec look good. But they also do with TMPG and CCE. And those are the mainstream encoders, which files will play in any SAP. As long as libavcodec has it's problems (ratecontrol), it just won't cut it.

http://www.licensing.philips.com/ser...ments1051.html

Use just for evaluation. Anyway test complete DVD is very very very slow ... perhaps more than complete reencoding. For me useless because I always know that my mencoder encoding was DVD compliant. At this time DVD-Video Verifier v1.5.0 is unintalled ...


Work perfectly on my SAP (Philips DVP-720, H&B AV512, and Siemssen SCO 5000 ND), Work perfectly on all other SAP tested with complete DVD encoding. I can make mencoder stream purpose testing on DVD forum and post complete compatible SAP list here if you want.


All my MPEG2 "low bitrate" encoding are always make with mencoder. I like to make MPEG2 encoding for 8 cm DVD. I make recently Star Wars Ripp with mencoder without problem.

No advanced test because I use only mencoder for low bitrate encoding ... but it's possible to make 5 Mbps encoding with mencoder.


http://www.licensing.philips.com/ser...ments1051.html


Use mencoder for make super-bit DVD is useless. I use Mencoder just for very hard encoding in very hard condition.

Well, you didn't answer all my questions, but I think the keyword is "low-bitrate".

I think you're just lucky, when you encode at low-bitrate, the spikes are probably less high. It's just a timebomb.

BTW, HCenc gives nice results also, and hasn't got the ratecontrol problem.

I just tried and IT IS NOT POSSIBLE to use the tool without registering it first. :roll:

Yes I suspected that. I guess we will have to make conclusions about the further credibility of sagittaire.
If he misleads us with this, I wouldn't rule out hasseling with testresults because "he is always right" :?

min available quantizer is in fact 2 for mencoder (minor bug in mencoder with x2 scale quantizer in mencoder) with custom matrix always > 16. For good security with RC you must target more than average q4 encoding with mencoder. Make "high bitrate (low average quantizer)" with mencoder is in practice useless and very difficult because:
- MPEG2 Q4 is for me high quality encoding. For example CQ 80 in TMPGEnc make encoding at ~ quantizer 4
- I use only Mencoder for make MPEG2 encoding in hardest conditions not for make super-bit DVD.
- Make "high bitrate" is certainely not the KDVD profil objective too


Quick Time Trailer are from my experience very hardest than DVD complexity. It's easy for me to make DVD ripp with mencoder but not with these trailers (more than 3 pass are sometimes necessary)

Choose some source here
http://www.apple.com/trailers/

And try to make mencoder encoding without underflow. I can make DVD compliant encodings for all these trailer (with equivalent bitrate < CQ 80 for TMPGEnc)
http://multimediacom.free.fr/Video/MENCODER.rar

And my english is very bad too ... don't forget that for my credibility ... :roll:

It's perhaps not legal but if I want log (and particullary 500$ log) I will try it always before. I try it for some sample. Too slow for complete check and DVD-Video Verifier v1.5.0 say only something that always know: "mencoder can done DVD compliant stream". DVD-Video Verifier v1.5.0 is unintalled now. But my english or licence for this log is not problem here.

I always just say:
Mencoder can make DVD compliant stream

You say:
Mencoder can't make DVD compliant stream. DVD-Video Verifier v1.5.0 prove it

I say:
Mencoder can make DVD compliant stream. DVD-Video Verifier v1.5.0 prove it

If you can't make compliant stream with mencoder it's simply because you don't use the good CLI. Upload your sample. I upload encoding with mencoder profil CLI for reproduce the encoding ... :idea:

For example only the third pass in my CLI done compliant stream for some Quick Time Trailer:

Mencoder can make compliant stream like I prove in this thread.

My mencoder HPII trailer (in first thread for test codec) is compliant:
- IfoEdit check say "no underflow"
- Mencoder check say "no underflow"
- hank315 (HC developper) say "best of my knowledge no error, stream is compliant"
- All my mencoder stream are good for all the SAP tested
- and now DVD-Video Verifier say "no fatal error for MPEG2 stream"

Anyway subject is close for me. Try to find incompatible SAP with this hard example and speak after about that ... :evil:
http://multimediacom.free.fr/Video/MENCODER.rar

Well-well, look at what we have here :).
So Philips DVD Verifier needs registration...
Very interesting!
Let me tell you a story.
Once upon a time there was a thread in this forum about bitrates and about how they were measured by BitrateViewer.
Don't ask me where the thread is because I just can't find it anymore.
So, in this thread everybody was using bitrateviewer to draw conclusions on how encoders were better or worse than other encoders.
The interesting part of it was that even using the same sources as some people that posted on that thread, using the same scripts, plugins versions, encoder versions, etc...I could never reach the same results as other people that even posted screenshots of bitrateviewer.
A few weeks past that, I went to visit my cousin and he told me how he found kvcd so useful and that he had done some encoder comparison to choose which one to keep installed.
Guess what tool he used to make the comparison? Yep, bitrateviewer!
Now the funny part of this.
He had found similar results to the ones found by other members of this forum that were completely different from mine 8O.
I spent hours at his place trying to figure out how it was possible, until I decided to reinstall all his tools.
And :!: bang :!: suddenly the results on his computer were very similar to the ones on my home PC.
And that's when he told me bitrateviewer was missing some bits of information.
I told him that bitrateviewer free version doesn't show all possible info and he laughed while he applied a crack he had downloaded days before for bitrateviewer.
And indeed there was the difference 8O.
Bitrateviewer free edition displays different results when compared with a cracked edition!
Which one is the more reliable?
Beats me!, but I'll never know because I don't use warez.
End of story.
Now, I'm not saying that Saggitaire used warez to show us his findings, but if by any chance he did, :arrow: then his results might be completely different from the reality ;).
Cheers

I don't need to go find one that doesn't work with Mencoder's encodes.
I have one right here with me that doesn't read mencoder's encodes properly.
It's and "el cheapo" AG DVD-336K.
No matter what settings I use with mencoder it can NEVER read it properly.
There! Satisfied?

I already knew that and, guess what, I discovered that... while testing mencoder :-) Perhaps it's a wanted trap put there by the author :D

well it's simple ... gamma can check directly my mencoder file with DVD-Video Verifier v1.5.0 and post result here
http://multimediacom.free.fr/Video/MENCODER.rar

I think that result will be exactly the same ... ???


1) You test this file ... ???
http://multimediacom.free.fr/Video/MENCODER.rar
(sorry it's slow ... I know)

problem here is not your mencoder encoding (gamma encoding are not compliant for example) but my mencoder encoding. I use very special multipass CLI for mencoder. Generaly mencoder make compliant stream only for third pass in most case ...

2) I can't find reference for this SAP. Certainely specific clone product. You know the chip used for hardware decoding ... ???

I quote: "Generally....most case....."

We need an encoder that produces compliant streams, period.

If an encoder "generally" produces an playable stream which in "most case" is compliant, then we don't need this encoder.

IF you found a way to encode a clip using mencoder with a special CLI and so on, which is really compatible, then it's great. For fun. Not for encoding video. Because of the simple reason that it's just not reliable, as you suggested yourself in your previous posts.

I'll try to take your sample to my work and verify it.

Well actually mencoder has simply difficult for Rate Control. If you exclude RC problem then MPEG2 structure stream is DVD compliant.

Make always compliant stream with one pass is actually simply impossible:
- vqscale mode is real constant quantizer mode : mencoder don't use vbv specification for make stream
- vbitrate mode use partially vbv specification : priority is not to respect vbv specification but target bitrate

Mean simply that Multipass to respect vbv is necessary.

- One pass bitrate can make vbv compliant but it's generaly not the case. Quality is crap
- Two pass make generaly vbv compliant stream for source like DVD. Quality is not optimal.
- For hardest source like Quick Time Trailer three pass are necessary. Necessary too if you use high bitrate for DVD source. Quality is optimal

I make always my encoding in three pass for highest possible quality. If you want know if your stream is compliant (at first, second or three pass) then it's very simple : Mencoder check always underflow problem.

That's what we're saying all along.
Glad you finally agree.

@Sagittaire,
Hey pal, I used to love mencoder and I was a strong supporter, as you all can recall.
But...either you tell us how to produce such a compliant stream with mencoder, or I'm affraid we won't be much supportive of it.
And no, I still haven't tested your clip because I haven't had time.
The DVD player I told you about is surely a clone of some other player out there, but I can't help you much with that.
All I can say is that such SAP used to be very popular here in Portugal due to it's price.
Cheers

Well I will make faq for my complete and detailled encoding procedure with mencoder (next week I think)

You test that and you say if my procedure work for you. I my procedure don't work for you then discution is closed and I change my signature ... :wink:

:lol: sounds fair :lol:

Lavcodec uses a LINEAR Quantisation where TmpgEnc, CCE and others do use a NONLINEAR Quantisation. So a Quantizer of 4 in lavcodec results in a significant worse output than compared to a Quantizer of 4 when using TmpGenc, CCE etc.!

Do make an encode using Mencoder where at very low lit underwaterscenes when Q gets kept at Q2 (vmin=2) recognisable 4x4 up to 8x8 Blocks do occur when surfaces do fade from dark blue to very dark blue or almost black. I almost solved that by using vmin=1 and tweaking the lmin argument which caused th Q dropping almost to 1.5 -> blocks where off, ... but on the other side this caused other problems in the ratecontrol.
I dont know if that issue has been fixed already or if its not the case when using your CLI arguments.

TmpGenc uses variable Q's when doing a CQ based encoding. Do watch the curve it still bahaves variable. This results in a for human eyes more good looking picture. Same in Case of OPV in CCE.

Really ? I did not notice that. The non-linear quatization is one of the most interresting enhancement of MPEG2 over MPEG1. It's stupid to not use it.

Do look in Bitrateviewer (jupp its mentioned in the mpeg header). Lavcodec resulted streams are generaly Nonlinear.

A Q of 2 in Lavcodec matches almost a Q of 4 in TmpgEnc, CCE etc. Keep in mind Im talking about human eyes subjective comparison.

Both types of quantisation do got their pros and cons. I think Kika could explain that very deeply. Anyhow I remember that Petercheat sometime offered to switch Lavcodec in Nuenc to nonlinear quantisation. But in his words it wasn't worth it imho. I dont remember correctly.

Do make an encode in TmpgEnc and check the quantisation, it will also be here a nonlinear one. Surprising. Out of Specs? Boooom :)

Said objectively:
I still dont see the sense of that war in here. The main spirit of KVCD was to NOT follow the specs, enlarging GOPs, switching, tweaking etc.
Thats also the reason why a SAP compatibility chart is given in the portal.
If the bitrate spiking of libavcodec is solved by some tricky arguments in the command line, ok. The rest, well ..... same in case of KVCD: If the SAP doesnt support it. Switch over or get a new SAP which is proofed playing back even DVDs of Oil where fries have been boiled in :)

I do expect from commercal DVDs that they are FULL specs compilant, the backups should be played back in my own SAP. If I have to handle a Job I dont make any experiencies and choose CCE or TmpgEnc.



Think about it anyway. That could be a pleasure in general, not only depending on all this lavcodec hadaik!

Linear :lol:

Note: I don't think that any spec specifies if the quantization has to be linear or not, but I may be wrong.

I don't see a point in using an encoder that you constantly have to tweak in order to keep things compliant. Keeping compliancy should be the encoder's problem, not mine.

I'm not allowed to use high bitrates and am supposed to take care that the avg quantizer is over 4? Yeah right..

IMHO, the main thing about compliancy is simply the bitrate spikes because of the physical boundaries. Long GOPs are not a (big) problem.

Hmmm, I'm affraid there's more than that (bitrate spikes) to it.
I once ran a couple of tests between Tmpgenc vs Mencoder.
I still have those KDVDs with me.
Yesterday I ran those mpegs through bitrate viewer and on one of the clips the bitrate never went above 4000 but still it doesn't play on my old SAP.
So I believe that bitrate spikes are not the only problem here.
BTW, I muxed those mpegs with Tmpgenc DVD author with audio 48Khz and I don't recall if it reported any errors during the authoring.

What comes to error reporting, I wouldn't trust TMPGEnc DVD Author a bit :wink:

Demux the VOBs and try muxing in MuxMan. It produces a very good analyser log if anything goes wrong.

Today I tested Sagittaire's sample. The VOB produced this:


After demuxing the m2v produced this:

Certainly looks alot better than my encoding with lavcodec :lol:

That brings us back to the point that I (and Boulder) stated earlier: an encoder should produce compliant stream without constantly tweaking it.

Ooops, wait a minute Gamma.
If Sagittaire's MPEG file reads a lot of errors on DVD Verifier but the demuxed M2V doesn't, then I would say that those errors were introduced by the authoring tool which doesn't make much sense...

@Sagittaire,
What authoring tool did you use on this sample?