FFmpeg vs FFvfw vs Mencoder ? - Page 22

digitall.doc · #**421** 04-04-2004, 05:27 PM

incredible,
sorry me if I'm wrong (I'll sure be), but in the graphis, each line means 1 Q value, but 1000 Kb. So, when bitrate drops below 1000 Kb, Q should be 0 to be under... I think you're right that we can use this visual aspect of Q under bitrate as a general guide, but I also think Q can go over bitrate, without meaning bad quality. Again, I'll be sure wrong...

Anyway, if you read my last test result, for vqmin=2, I got almost always Q below bitrate with lmin=1 or 1.5. And they were my best results.

bilu · #**422** 04-04-2004, 05:28 PM

Quote:

Originally Posted by digitall.doc

I understood you want to fing the minimum bitrate needed to get acceptable quality... but you will find it in the context of many other settings (with which we feed mencoder), so you won't really know if the effect you get is really just due to bitrate...

It doesn't work like that

My testing samples have been small (between 2 and 7 minutes) so the average bitrate is meaningless here. It's the resulting bitrate and quality of the tested settings that matters to me

Bilu

digitall.doc · #**423** 04-04-2004, 05:30 PM

Quote:

Originally Posted by bilu

About your tests I had made them all, just didn't keep records.

In this point of testing the video would speak louder than filesizes or average bitrates, it would be better to upload videos than those values.

You already did them?, I wasted my time

And, yes, you're right, better see the resulting videos. But if I cannot upload a single image capture, imagine a video sample...

bilu · #**424** 04-04-2004, 05:32 PM

Quote:

Originally Posted by digitall.doc

vqmin=2:lmin=2
==============
File size: 148882 Kb
Bitrate: max 5165 avg 2090
Q value: max 4.08 avg 3.21
Q graphic: it keeps around 3, but seems to raise easily when needed.

vqmin=2:lmin=2.5
==============
File size: 115904 Kb
Bitrate: max 4093 avg 1627
Q value: max 5.26 avg 4.29
Q graphic: varies between 4 and 5, with tendency around 4.

You'll see that vqmin=2:lmin=2.49 still behaves like lmin=2.
But with vqmin=1 it will look better and the filesize won't grow

I imagine that using p_mask=1 it will look even better

Bilu

incredible · #**425** 04-04-2004, 05:36 PM

Quote:

Originally Posted by digitall.doc

incredible,
sorry me if I'm wrong (I'll sure be), but in the graphis, each line means 1 Q value, but 1000 Kb. So, when bitrate drops below 1000 Kb, Q should be 0 to be under... I think you're right that we can use this visual aspect of Q under bitrate as a general guide, but I also think Q can go over bitrate, without meaning bad quality. Again, I'll be sure wrong...

Anyway, if you read my last test result, for vqmin=2, I got almost always Q below bitrate with lmin=1 or 1.5. And they were my best results.

Q should not only base on the low bitrate count as you said like "1000kbit"!
IF the bitrate goes very low and I do set a determined "border" using VQmin=2 ... then the quantization is much too high for these underwaterscenes as it will be kept at 2!, means there WILL result blocks (in all my tests).. So as you can see in most parts of the graphic its working right, means the lower bitrate also lets drop the Q to about 1.5 which is really recognisable!

audioslave · #**426** 04-04-2004, 05:38 PM

Hi guys!
Has anyone heard anything about the re-writing of the bitrate control yet? (It was the rate control they were gonna change, right?

) Looking good fellas. Very interesting info you're sharing with us here!

@bilu
Do you have a sample we could download - just to see the quality? I'm VERY curious!

bilu · #**427** 04-04-2004, 05:41 PM

Quote:

Originally Posted by audioslave

@bilu
Looking good. Very interesting info you're sharing with us here! BTW Do you have a sample we could download - just to see the quality?

Not at home right now

And I don't have the space for hosting it

Why don't you try my settings on a sample?

NOTE: my settings are not quality-oriented but "acceptable quality vs bitrate" oriented

Bilu

audioslave · #**428** 04-04-2004, 05:46 PM

@bilu
I sure will try. I haven't had any luck with getting MEncoder to work so far

. I'm no computer wiz but I'll keep on trying. I'll even rip the intro scene from The Abyss to use for experimental sample!

Newbe alert

I just copy the bat and ini steeings you posted earlier, right? I must be missing something since everybody else seems to be able to use MEncoder... What compile do you recommend for me to download?

bilu · #**429** 04-04-2004, 05:54 PM

Download this one:

http://clientes.netvisao.pt/bilu/bru...er20040403.zip

Bilu

audioslave · #**430** 04-04-2004, 06:02 PM

Thanks. Downloading right now...

About the *.bat and *.ini files:
What setting shall I use here? Please forgive my ignorance but as I said earlier; I haven't had much luck so far - with MEncoder that is.
Sorry to be a pain in the -BEEP-

audioslave · #**431** 04-04-2004, 06:37 PM

Hmm, no luck yet...
This is what I used:

abyss.bat

Code:

mencoder -include settings.ini -vf-pre softpulldown -lavcopts keyint=18 movie.avi -o movie.m2v

I have my fake avi in the same folder as MEncoder

settings.ini

Code:

of=rawvideo=1
ovc=lavc=1
nosound=1
noskip=1
vf=yuvcsp
lavcopts=vcodec=mpeg2video:vrc_buf_size=1835:preme=2
:precmp=2:ildct=1:ilme=1:vstrict=-1:autoaspect=1:vrc_minrate=300
:vrc_maxrate=9800:vqblur=0:vqmin=1:mbqmin=1:lmin=1
:intra_matrix=8,9,12,22,26,27,29,34,9,10,14,26,27,29,34,37,12,14,18,27,
29,34,37,38,22,26,27,31,36,37,38,40,26,27,29,36,39,38,40,48,27,29,34,
37,38,40,48,58,29,34,37,38,40,48,58,69,34,37,38,40,48,58,69,79
:inter_matrix=16,18,20,22,24,26,28,30,18,20,22,24,26,28,30,32,20,22,24,
26,28,30,32,34,22,24,26,30,32,32,34,36,24,26,28,32,34,34,36,38,26,28,30,
32,34,36,38,40,28,30,32,34,36,38,42,42,30,32,34,36,38,40,42,44
:vbitrate=3000:vqcomp=1:vqmin=1:mbqmin=1:lmin=2.49:scplx_mask=0.24

Do I need to delete the linebreaks somewhere in settings.ini?

When I start the abyss.bat a DOS window pops up and a bunch of text scrolls down then the windows disappears. It goes by so fast I don't have a chance to read anything in the DOS window... What am I doing wrong?

I really wanna test this encoder, believe me.

EDIT: I managed to see something about a config file that couldn't be found but I don't know were the file is supposed to be located or what kind of file it is...

vmesquita · #**432** 04-04-2004, 06:37 PM

Ok, I finished my encode.

The idea of using temporal noise for AVI was not good, the temporal noise became even worse dancing blocks. I am trying now with gaussian noise, looks much better. Also for AVIs, the internal post-processing looks very good, much better than denoise3DHQ (to remove DCT blocks).

@digitall.doc
Looks like vbitrate=9800 makes the encoder use more bitrate, but I still don't know the effects quality-wise... Very good idea using lmin to make a CQ like prediction.

audioslave · #**433** 04-05-2004, 06:00 AM

Oh, man... This was really complicated. I have read the newbie guide to MEncoder but guess what - I still can't make it work! Could you gurus please help me finding out what's missing? I really hope so. Here we go;

1) I downloaded MEncoder from the link bilu posted - "mencoder20040403.zip" - and unpacked it to "D:\DVDRip\MEncoder".
2) I downloaded "mencoder_avs.zip" and unpacked it to "D:\DVDRip\MEncoder" - same folder as MEncoder.
3) I copied and pasted bilu's settings.ini and saved it as "settings.ini" to "D:\DVDRip\MEncoder".
4) I have made a fake avi and saved it as "video.avi" to "D:\DVDRip\MEncoder" (It's a clip from the NTSC version of "The Abyss").
5) I used this for *.bat:

Code:

mencoder -include settings.ini -vf-pre softpulldown -lavcopts keyint=18 video.avi -o video.m2v

No line breaks...
6) When I run the *.bat file a DOS window pops up and then immediately closes. There's text in the window but it all happens so fast I can't see what it says...

I hope this is all the info needed to be able to solve my problem?
Thankful for help.

bilu · #**434** 04-05-2004, 06:27 AM

Open a DOS window, go to that directory and run that command.

This way we can at least know what is the error message

Bilu

incredible · #**435** 04-05-2004, 07:41 AM

@ Audioslave
Add the mecoder Path to your systems environment variables in the system settings window of windows.

Explained here:
http://www.kvcd.net/forum/viewtopic.php?t=9093&start=24

digitall.doc · #**436** 04-05-2004, 11:21 AM

Quote:

Originally Posted by vmesquita

Looks like vbitrate=9800 makes the encoder use more bitrate, but I still don't know the effects quality-wise...

Well vmesquita, I don't know if this is the right way, making mencoder use every bitrate is needed without a limits (in fact limited to 9800). I think bilu is going the other way, and sure he is right. But this approach is giving me very good results for KDVD encoding. Of course, when talking about SKVCD or KVCD, we should go other way, but for KDVD...
I have a doubt: if we want mencoder output better quality, do we set lower quantizer values?. With lower quantizer values, what will the tendency of the encoder, to raise or lower bitrate?. Thinking on this is why I let mencoder raise whatever it wants bitrates, in order to keep as much quality as possible. But, again, I'm no expert in this matter, and I'm sure wrong. But didn't find yet a better explanation.

Quote:

Originally Posted by vmesquita

Very good idea using lmin to make a CQ like prediction.

I'm happy I got a reaction to my tests!. I was begining to think it wasn't of interest for anybody

. As you see, keeping the rest of parameters without change, I did get different final filesizes and quality files. For a vob that was 468350 Kb I got a file size range from 257376 to 103841... but lmin 2.5 and 3 wasn't that good, and maybe would need some changes else. I still think that, once we get a set of good standard parameters for KDVD, we can adjust file size through lmin.

@incredible,
I'll redo the tests I did, but now with vqmin=1 and lmin between 0.5 and 2 or 2.5 to see the results. I agree we should let mencoder the opportunity to go down to 1 when needed, but I don't know the way to do that, since the majority of times a Q value of 2 will be enough. But I'll do tests with vqmin=1 and see quality and filesizes.

@bilu,

Quote:

You'll see that vqmin=2:lmin=2.49 still behaves like lmin=2.

But it didn't, bilu, my friend. lmin=2 looked better than lmin=2.5, and related to filesizes, bitrate and Q value, you saw my results...

I'm begining to worry that with different compilations, we are getting different results...

incredible · #**437** 04-05-2004, 12:04 PM

Well Digi.Doc ...

As I said EVEN when using the build I do use ... at Vqmin=2 blocks on very low bitrates liek underwaterscenes do appear as "maybe" in mencoderspeach a quantisation of 2 already does its job too deep

Another approach could be a new Matrix where low frequencies will be almost "untouched" means veeeery less filtered.
Im very shure that mencoder does cry out for a special "mencoder egoistic" matrix as he behaves totally different then TmpgEnc or CCE!
He does Quantize in average veeeery less (seen as values) but at very low bitrate scenes he likes to punch the quantisation (no matter if Vqmin is set to 1).

But you have to say mencoder that he has on exactly these scenes to lower the quantisation ... which works as you see in my pics above, not continously but its an approach. And exactly thats what (to me) makes mencoder amazing.

Sorry if I didn't test your parameters till now, cause with my settings I can just use Vbitrate to determine (almost) final average bitrate ... and in 80% of cases it really matches

... bu anyway ... a prediction is obligatory ... as Vmesquita also mentioned that in here as NO encoder knows what will happen in further treatments of the movie

But as I said, this Weekend I got my first movie where a Bitrate Peak-Juuuump happened! So my settings wheren't optimal ... yesterday I did set the Q between 1-8 (before 1-3) and a Vmax_bitrate used at 8000 I did change to 5000kbit, .... today I will let encode that stream again and lets see how it will behave.
For me its logic ... the encoder did encounter a very complex scene, the reaction delay got crazy and only was to be allowed to rise Q till 3 means therefore a Veeeeery high bitrate resulted to keep that Q edge of 3.

Well I do my "K19" Encoding again as this seems to be a very "sensible" one for mencoders engine.

According to quality still I do keep my old version of mencoder but on the other hand I cant encode in m2v directly without that nonsense space needed demuxing way

bilu · #**438** 04-05-2004, 12:06 PM

Quote:

Originally Posted by digitall.doc

@bilu,

Quote:

You'll see that vqmin=2:lmin=2.49 still behaves like lmin=2.

But it didn't, bilu, my friend. lmin=2 looked better than lmin=2.5, and related to filesizes, bitrate and Q value, you saw my results...

Try 2.49 and you'll see that the behaviour is much closer to 2 than to 2.5

Bilu

incredible · #**439** 04-05-2004, 12:11 PM

It would be veeery interesting if people do tweak lmin etc HOW that results in the relation from Q to bitrate! Means could you also post a pic from Bitrateviewer added by a quote of the commandline or parameters used??

Thanks in advance a lot!

Webspace does exist at lycos or wherever ... no matter if advertising added, it should at least have the permission to link to jpegs/Gifs so they can be posted/called in/from here.
Everything else is just assuming ... and the Bitrate-Pic postage would make develeopemend faster

digitall.doc · #**440** 04-05-2004, 04:58 PM

Some results else from a new test. This time I tested incredible/bilu way with vqmin=1, and tested several lmin values.

Mencoder dev-CVS-040401-06:00-3.3.1 build, compiled with Cygwin.
The sample is the 4th vob of Matrix Reloaded film (it lasts about 9:29 min). Original file size: 468350 Kb Again no filter, just crop and expand.
Command-line and settings as posted before. Encoded at 9-11 fps.

vqmin=1:lmin=0.5
=============
File size: 497146 Kb (bigger than the vob

)
Bitrate: max 10254 avg 6986
Q value: max 1.88 avg 1.08

vqmin=1:lmin=1
==============
File size: 285186 Kb
Bitrate: max 8661 avg 4006
Q value: max 2.40 avg 1.79

vqmin=1:lmin=1.51
==============
File size: 177093 Kb
Bitrate: max 6063 avg 2487
Q value: max 3.31 avg 2.83

vqmin=1:lmin=1.52
==============
File size: 176753 Kb
Bitrate: max 6045 avg 2482
Q value: max 3.31 avg 2.84

vqmin=1:lmin=2
==============
File size: 148882 Kb
Bitrate: max 5165 avg 2090
Q value: max 4.08 avg 3.21

vqmin=1:lmin=2.5
===============
File size: 115904 Kb
Bitrate: max 4093 avg 1627
Q value: max 5.26 avg 4.29

Related to filesizes, 1:1 gave a big file. I'm sure that you look at the figures you get, but with this filesize I won't be able to fit 2 films in a DVD-R (not to mention putting 2 audio streams...). 1:2 gave the same filesize and bitrate/Q values as did vqmin=2:lmin=2 (I think it was bilu said it happens), and maybe is closer to desired final file size. But I don't see the advantage of using vqmin=2 with lmin=2 (Q value are always above 2).
I tested with lmin=1.51 vs lmin=1.52 to see if is there any difference... and there is: file size, in this small sample, decreased 0.2%... that means this is a way to adjust final size, since from lmin=1 to lmin=2 there's a 48% file size decrease, and lmin=2 is still very good. Again, I think that this can be the way to adjust final file size, since with 1pass CQ-VBR we cannot make a better file size prediction.

incredible, I still have to give your matrix a test. When you talk about a specific matrix for mencoder, could be yours the matrix

. I'll test.