What VirtualDub version do you use? VirtualDubMod?
 

Hi guys,

I have a laptop PIII-500/ 192 RAM / 2GB free HD that I can use for Mencoder tests. All my contributions were theoretical so far and now it's time to come into action! :twisted:

But I need a hand to "synchronize" with your tools and testing methods:

- What tools do you use besides Bitrate Viewer? And are you using the free version?
- What VirtualDub version do you use? VirtualDubMod?
- Something else that you'd like to point out?

I'd also like to be pointed to common settings used in TMPGEnc and CCE since I never used them before and it may help me establish comparisons.
Please understand that my previous experience is MPEG-4 and not MPEG-2. :oops:

Thanks in advance,
Bilu

Hi Bruno,
I'm using mencoder from the official site MEncoder 1.0pre3-3.2.3.
Get it here ftp://ftp.mplayerhq.hu/MPlayer/relea...32-1.0pre3.zip
Along with ffvfw-20041301.exe by Athos from doom9 (you know where to find).
Then I use a codec.conf file that was posted by VMesquita and that enable you to use fake AVI done with makeAVIS that comes with ffvfw.
Get it here http://www.jltoca.uaivip.com.br/files/mencoder_avs.zip
Inside the zip by VMesquita there is something else to enable frameserving via VDubMod. I use v1.5.10.1b2424 but haven't tried it with mencoder, only with ffvfw.
Oh, Avisynth 2.54 with MA script by Kwag with a small touch by Inc :)
I use Bitrate Viewer. Only have the free version and it does the job.
About CCE and TMPG I always used the templates that you could download at the site but if u need something just ask :wink:
Cheers

Thanks Rui :)

I'll avoid the Windows-specific parts since my final objective is Linux/FreeBSD encoding.

So I won't use:
- ffvfw (until I know mencoder better and be able to establish similarities)
- Avisynth (most of my Doom9 post were Avisynth-related... :roll: )
- MakeAVIs - I'll work mainly with VOB samples;

But this will also motivate me to explore mplayer/mencoder filtering :)

About CCE and TMPGEnc, I'll check the templates ***IF HUMAN READABLE***. I won't install CCE or TMPGEnc just to be able to load a binary template. My purpose is to see what is commonly accepted, it will bring me a better perspective on MPEG-2 encoding and I'll be able to compare parameters better.

Thanks again :)

Bilu

Hi Bruno,
From what I recall CCE templates by VMesquita are regedit patches and TMPG templates are a bit confuse.
In need of help just honk and I'll try to aid you.
Be aware that using mencoder with VOB files and without avs scripts is going to enable ~7fps which is "far out dude" for a PC with similar specs to mine that is providing me with 4fps with current avs script.
I'm saying this if you aim to DVD res 704x576 else you'll have better results :wink:
Cheers

Hi bilu (Bruno :wink: ),
so you're leaving your "coach" tasks and are begining to play with us: welcome to the match!.
I think you already got very good help from Rui. Nothing to add, even less since I work on Windows and you'll on Linux/FreeBSD.
About mencoder parameters, you already know my good results with vrc_..=tex, vqsquish=1, trell, cbp,... and tweaking with B frames. I recommend you to employ B frames to mantain compatibility (even I didn't test a encoding without B frames in my standalone)
If you want to do KDVD, and try to fit 2-3 films in a media, do try Notch matrix, since it will decrease filesize keeping good quality.
We're all waiting your results 8O
:lol:

@rds_correia

With vqblur=0 and vrc_eq=tex, without any other filters or settings, I was able to achieve 16 fps :!: :!: :!: on my PIII-500.

Results so far:

1) Interlaced encoding doesn't work. Maybe when parameters like ilme (interlaced motion estimation) get ported to the Windows version :roll:
2) Best IVTC filter combo was -vf softpulldown,ivtc=1 -ofps 23.976 and still ain't good enough on low-motion.

3) The documentation sucks (not really new) :roll:
From now on (at least until ilme gets ported to the Windows version) I'll use DVD2AVI Force Film over the NTSC FILM streams. I won't do tests with deinterlacers or Avisynth.

My tests will be done with progressive VOB files, PAL and NTSC. It's easier to unleash the speed of mencoder this way ;)


Bilu

@digitall.doc

My next task we'll be test the interesting parameters you've been testing until now ;)

And 2-pass too, of course :)


Bilu

Hi Bruno,
Post your command line ASAP :!: :!: :wink:
The rest I'll try to post in the evening. I'm at work now...
Cheers

Well, my friend, the result of your interesting advises
:)

Usage: encode.bat movie.vob

This command line was not very efficient but is fast!
Used with -vf %ivtc% for test deinterlacers and IVTC filters.

My company's laptop has WinXP, it's my testing platform for now.


Bilu

Hi Bilu,
:( Still at work :(
As soon as I get home I'll try one or two arguments from your command line.
BTW since you're in Alcochete :arrow: Portugal, why are you using keyint=18 instead of 15?
NTSC sources in PAL land or just trying longer GOP to achieve better compression?
Also, what do you mean by fast but not very efficient?
Cheers neighbour :wink:

I got lots of problematic samples from Doom9 users (remember that one of the most used Avisynth tool was the Telecide/Decimate combo, most MPEG-4 encodes are progressive and made from Telecined 24fps->30fps when NTSC).

Pure telecine, not problematic:
13.391.872 telecine1.vob - Monty Python
19.509.248 telecine2.vob - LOTR II intro

Problematic:
46.243.840 anime_telecine.vob - Read Or Die
35.921.920 hybrid.vob - Star Trek TNG hybrid scene
52.293.632 interlaced.vob - Neo Genesis Evangelion anime

Total NTSC samples:
5 File(s) 167.360.512 bytes

I also have AVI samples, but I work more with the VOB samples.

About PAL land I have about 10 DVDs, only one is problematic: The Abyss. Field-shifting (gets interlaced in some parts because of field dropping) and non-anamorphic encode of a 16/9 stream (I made an AVS script to allow an anamorphic reencode - a friend of mine did it with CCE for me - that looked much better than the original).

By fast and not very efficient I mean that it didn't use trell, cbp, mbd=2, mv0, or a denoiser like hqdn3d (heard good feedback about it) and the file got very big, relatively to one try with all those parameters.

But now I'll do Force Film in all my NTSC telecined samples and use some PAL ones too - not Abyss - to be able to use progressive VOBs as source.

Then I'll test each parameter I mentioned, it's impact in size and in speed.

All my tests will be

1-pass
vbitrate=vrc_maxrate=9800
vrc_eq=tex (same as CQ 100)
vqblur=0 (same as no quantizer delay)

So quality will allways be maximum and quality improvements can be measured also by the filesize - macroblocks describe the movie better with less need for compensations. With motion estimation and rate distortion parameters, of course. And that's what I'll test first.


Bilu