Optimizing Xvid for TV Output (summary)

[marnum]

Hi, this is my first post. May I introduce myself: marnum from Austria.

I've been reading this forum for some weeks because I've been very unsatisfied with the results Divx gave me. I like watching my movies on my TV-set, and I've encoded my movies with Mpeg2 for a long time. Now I found a TV cable that is 15 meters long and thus I can connect my PC to my TV-set.

The preprocessing options for TV output I've found here are:

Output from DVD2AVI with TV scale

Convolution3d or TemporalSmoother for removing invisible picture information

YUY 2 output (from the lumi-masking thread) to remove blocks in dark scenes
or
Levels

The encoder settings:

Lumi-Masking in both passes (You can do it if you use YUY2 output)

Raising min quantizer to 3 (The picture failures that occur now can't be seen on a TV-set)

Lowering max quantizer (I don't know how far I can go with that without losing filesize accuracy)

Finally, playback settings on TV-set:

Software for TV-output: TVTool 6 & ZoomPlayer or WMP 6.4
ffdshow filter with noise addition (for 1CD-rips) and postprocessing


Is there some option that I've forgotten?

[iago]

@marnum

I really can't understand how you have drawn "these" conclusions as a result of your reading the threads here, and I don't know which one to begin with ;). Still, just my humble opinion on some of your settings:

* In DVD2AVI, PC scale has always given better results than TV scale for me so far, though I have also tried TV scale in my tests to see if it works, but unfortunately with no success or no better results than PC scale.

* Denoising filters don't seem to be much relevant here; well, that is, it's not something "especially" related with TV output. You can simply use them or not.

* Levels seems not to help much with the issue actually. Instead, try lumoff=-2 with MPEG2Dec3.dll.

* As for lumi masking, I don't use it any more and don't suggest it to anyone either for PC or TV viewing oriented encodes. However, Koepi lately remarked that he gets good results without problems with it (together with lumoff=-2 and some smoothing).

* Raising min quantizer to 3 is "absolutely" not a good idea, keep it at 2 for both I and P frames.

* As for the max quantizer, you can use 6 for I frames and 12 or 16 for P frames imo. XviD is almost perfect regarding filesize accuracy and doesn't lose it easily ;).

* With ffdshow, use "mplayer noise", which adds noise to the dark areas as well. As for post-processing, in my experience it doesn't improve picture quality, but on the opposite degrades it.

regards,
iago


edit: Oh, Koepi is here too, replying simultaneously with me as usual ;).

[Koepi]

EDIT: wow, iago was faster again ;)

But for your initial post, marnum, here's my oppinion:

Ouch, this is a quite inaccurate summary which I can't "back-up" like it is. A summary would be more like this:

- Create your usual d2v file with dvd2avi, set it to YUV color space.
- Use MarcFD's mpeg2dec.dll with postprocessing (fetch it here (http://ziquash.chez.tiscali.fr/)) with avisynth (LoadPlugin="...")
- Modify your MPEG2Source() line and add lumaoff=-2 (should look like this: Mpeg2Source("MyMovie.d2v",lumaoff=-2)
- in high bitrate scenarios, use Trbarry's unfilter.dll (fetch it here (http://www.trbarry.com/UnFilter.zip)) and load it in your AVS
- Add a line like unfilter(5,5) which sharpens the image a little, amplifying the noise (important for the dark areas)

With this done, you're nearly finished. If you're low on bitrate, you can use lumimasking for encoding, but it can give slight worse results than without it - for me it looks good anyway.

Capping the quantizers at min=3 isn't a good idea, give the codec some space to play with. Really.

TV Scale in dvd2avi is just working in RGB color space btw, it's useless in YUV color space - but without avisynth, using vfapi (thus staying in RGB) it does the trick as well.

I hope this helps,

Regards
Koepi

[marnum]

First, thanks very much for your quick answers.

>> I really can't understand how you have drawn "these" conclusions as a result of your reading the threads here, and I don't know which one to begin with

Well, there are so many pro's around here with so much experience and they all post it in here - maybe I've picked up the wrong parts, but I

>> Denoising filters don't seem to be much relevant here; well, that is, it's not something "especially" related with TV output. You can simply use them or not.

I think that I'll get better quality if I use denoising filters (generally, not only for playback on TV).

>> Raising min quantizer to 3 is "absolutely" not a good idea, keep it at 2 for both I and P frames.

I got that from DivX and I thought it would also apply to Xvid.

One last thing: What do lumoff & unfilter actually do?

[iago]

->One last thing: What do lumoff & unfilter actually do?

Helps you get rid of the most dreadful black-blocking (macroblocks in dark/black parts of the encode especially when viewing on TV) issue ;).

iago

[marnum]

Thanks for your answers.
I encoded "Road Trip" for
1 CD (av. bitrate at about 990 kb/s) at 544x? (16:9 format)
using Xvid (latest Koepi version) and DivX Pro to compare them.

Compressibiliy Check = 2,25

AVS-file settings used for both:

mpeg2dec3
lumoff=-2
Convolution3d(preset=lowqu) Preset for noisy DVD source
LanczosResize
unfilter(5,5)

Encoding settings Xvid:

min quantizers: 2 both
max I-frame: 6
max P-frame: 16
H361 both passes
no lumi-masking
everything else = default settings

2nd pass ext. calculated by GKnot using f-corr and motion-corr.

Encoding settings Divx 5.02 Pro:

min quantizer: 2
max quantizer: 10
GMC & Bidirectional used, no Qupel
PVE = light
RC averaging period = 7500
everything else = default settings

Result:
Xvid looks worse than Divx, less detail and more artefacts. Divx simply looks better, when I added noise it looked just like the DVD (or even better, because the DVD had some blackblocking sometimes). Maybe I've left out some vital settings, but I was quite disappointed by Xvid. Divx is easier to use and there ain't so many settings where you can do so much wrong.

[Koepi]

Well, read through the threads here how to optimize XviD.

If divx gives you better results then use it. Btw., what did you use for playback? At encoding, which search precision? aiming which filesize? Ah, and don't compare apples and eggs, if you use PVE in divx, use lumi masking in xvid (hugest obvios mistake from the few settings you provided).

Just saying "I'm dissapointed" while you setup the codec wrong is clearly not nice.

Thanks for further infos.

Koepi

[iago]

@marnum

First, it's almost impossible for XviD (without b-frames) to have "less" details than "DivX502Pro" and even more impossible when you use b-frames and psychovisuals with DivX.

Second, why do you use GKnot to scale the curve for second pass external, when there is Koepi's StatsReader to do this job greatly?

Third, there are many options to tweak in XviD, but it will take you some time and effort to read the threads and keep up with all the new features to set up the codec correctly for better way results. Shortly, yes, XviD is more difficult to use than DivX5.02Pro, but when you "can" use it, it's worth the effort!

Next, don't worry, there are many people "never disappointed" with XviD when comparing results with DivX. But if you are "so" disappointed, you can happily use the easier DivX5.02Pro.

regards,
iago

[Didée]

Originally posted by marnum
...
Encoding settings Xvid:
...
2nd pass ext. calculated by GKnot using f-corr and motion-corr.
...
Motion correction for XviD-stats? Very interesting!
I never tried that, but I could imagine that this procedure delivers cr*p!

Fill some diesel into a Porsche, and you will definetly be disappointed by the performance ;)

[yaz]

hi marnum & all

there's a lot i understand but there's a lot a don't

>1 CD (av. bitrate at about 990 kb/s) at 544x? (16:9 format)
pretty high bitrate, xvid should do something miraculous :-)

>using Xvid (latest Koepi version)
what's that xactly (latest stab or unstab,release date, ...)

>Compressibiliy Check = 2,25
what's this & how did u figure it out? is it ~0.44% compressibility, or what ?

>Convolution3d(preset=lowqu) Preset for noisy DVD source
is that dvd really so noisy? lowq can make pretty harsh smoothing

>unfilter(5,5)
after c3d(...lowq) it's quite unnecessary. (5,5) makes nothing.

>H361 both passes
i guess, u mean h263. it makes a kinda smoothing, especially the background can turn to quite flat. use this only if u have compressibility problems. (but u haven't, if i got the above parameter well:-)

>no lumi-masking
that's ok, but then don't use pve in divx5 neither. the same effect, but xvid does it much better (imho)

>everything else = default settings
... xactly ?

>2nd pass ext. calculated by GKnot using f-corr and motion-corr.
nononono ... gknot is for nandub, use it with care. i wouldn't suggest motion correction even there but it gives crap for xvid, definitely.

>Xvid looks worse than Divx, less detail and more artefacts. Divx simply looks better...
it may happen but very rarely. (say, '6th day' i burnt with divx. lotsa reasons but the only real point was that my friend wanted that :-)))

>Maybe I've left out some vital settings ...
i bet, u did :-)

>... but I was quite disappointed by Xvid ...
don't give up! it's much better than divx

>Divx is easier to use and there ain't so many settings where you can do so much wrong.
a question of viewpoints. the more settings u have the more faults u can do, right, but ... the more flexibility u have to make your encode perfect.

before u drop xvid away, give it some more

y

[marnum]

@ iago:

1. I won't use B-frames as long as they aren't implemented in the stable version.

2. I read that GKnot can build better stats-files. I thought that would give me better results. Thx for your correction.

3. See, I don't want to spend all my time with encoding movies, because I've got other things to do either. I'm seeking for a more or less fast, stable (I mean longer-lasting), and user-friendly method to "back up" movies. If Mpeg2 was better I'd use SVCD (Actually I use it for some movies e.g. Monty Python & the holy Grail).

@ yaz and all the others:

See 3. point

@ yaz especially:

Maybe you shouldn't pick up every line a person writes and then falsify everything. Some people may feel offended by such posts (I'm not, but some may).

[vinouz]

I mostly feel the same as marnum.
Most people will get bored (sot scared, bored) by the time it must take to get Xvid. But not because of it's huge number of options, but because you need to look 5 threads, filtering all the trolling, just to get how to use correctly one options. And on the respective influence of one option on the others...
I know it takes time and it's boring. But I really liked Koepi's Xvid explained. But now it's obsolete, and we definitely need some sort of summary of all the options somewhere. U don't need to discuss all, but just to specify simply what affects what and in what order. Then people are big enough to figure out what final effect it would give.
Did you try the personnalized settings on a car in GT3. I definitely appreciate it, mostly because of the concise and clear documentatoin on each setting. I bet it must take about 1k text per setting. Not more.
We all would like to tweak Xvid. But people need a little doc. Not sparse settings found a la va comme je te pousse on 25 threads.

And thanks for all the answers on this thread. I find this one pretty useful :)

vinouz

[UGAthecat]

Originally posted by vinouz
...I really liked Koepi's Xvid explained. But now it's obsolete, and we definitely need some sort of summary of all the options somewhere.

Maybe a wiki like the avisynth2 group has set up would work?

[Shayne]

I dont know if divx 502 is so much better and easier. Appears to me to be a bit more difficult to install.

Why? went it appears you have no idea how to operate XviD are you messing around with all these filters. You should walk before you run. Click the default button follow Dooms mini guide leave C3d Unfilter and all the rest alone and see what you get with the core codec. Once you impress yourself you can tweak it like the rest. read test and understand.

[yaz]

@vinouz/marnum &all i offended
shame on me if u feel that :-((( sorry boobs !!! anyway, i wonder which of my sentences was/were so much disturbing.
btw, why don't u presume that i ask so as to help u. if u don't answer that i (& anybody else) can't help anyway.
just some drops :
- it's impossible to write a comprehensive guide cus xvid is under a continuous & heavy development. options come&go, get in&out, changes all the way. the only way is to follow the threads. sorry, but that's what we have.
- before u've got the basics u can't go on. (before u start running u must learn to walk, befero u do that must learn to stand up. & u must let loose your mom's hand :-)
- ... (yeah, i'm getting to bore myself :-)

y

ps life is hard but that's why it's worth to live

ed: i've got somethin vinouz mentioned (short description&explanation of the options) i wrotre to my (lazy:-) friends. if there's any interest i can brush it up & drop it on.

[vinouz]

---------------------------
Edit : thanks iago, for you putting a guide - Sure it helps ! - so I hope my babbling being a bit obsolete now
for beginners following, it's on this thread (http://forum.doom9.org/showthread.php?s=&postid=202809#post202809)
---------------------------

Well, I didn't ask for a comprehensive guide. I ask for a summary.
Let me explain. I know how to walk. among other things, I used XVid last year, took time to see what its options gave, and did some 1 and 2CD rips. It took time, and yes, I have pleeeenty other things to do in my day than to rip&encode.
If man had to learn to walk on a SuperMario like moving ground, I think man wouldn't walk. (ouch! such a wonderful image ;oP)
Coming back, I find the situation even more messed up. (But yeah, the codec greatly improved ;)= ).
But I don't feel like everyone has to finger himself before encoding.

Ok, but let's stop trolling. I don't know what a wiki is (probably shame on me) - and I'm going to take a look at avisynth forum btw -
Just a little idea : why not a persistent thread with the first post indicating for each option, which thread(s) cover it. That would already be a big improve for ppl to learn Xvid.

Because I just can't imagine how many people like this don't dare to ask, don't have time to investigate, want to use Xvid because it's a wonderful codec, and tweak it nonsensely in many ways.

Thanks,
vinouz

[sffubs]

@UGAthecat: a wiki would be an excellent idea! The one that the geoshell project is using (http://geoshell.sourceforge.net/) is very useful and provides good documentation where there wasn't any before. In the very least it could have explanations of what exactly all the parameters do, and most of the stuff to go in it is already littered about the forums.

Cheers,

sffubs

[marnum]

@ yaz:

>>@vinouz/marnum &all i offended

I don't need a wise ass telling me that I've done everything wrong. Be happy god has blessed you with so much knowledge.

Edit: The 2-pass guide is very good and newbie-friendly ;-P - but why mustn't I change the Alt.curve parameters?

[marnum]

Has someone thought about writing a guide for optimizing for TV-output? One could also combine iago's 2pass guide (if he allows that) with 1 or 2 preprocessing avs scripts and some nice ffdshow and TVTool settings...

[Koepi]

@marnum:

a) no profanity, insults,... rule no. 4 is a rule which you should read carefully.

b) you have all hints in post 2 and 3 in this thread - why another guide? iago mentions that luma offset stuff in his "quick start 2pass guide", so it IS already covered.

EDIT:
c) marnum wrote:
Edit: The 2-pass guide is very good and newbie-friendly ;-P - but why mustn't I change the Alt.curve parameters?
Because it's hardcore stuff. The search function of this forum is really useful, we have a _huge_ thread about Alt.CC, you may want to read some posts in there and get an idea why Alt.CC is in advanced settings... Use the usual stuff and get good results. Mess around a little with AltCC and totally mess up your encode. You may tweak maybe 5% more quality out of AltCC, but you can degrade it 100% with that as well, so you have to experiment a lot with it to find and understand how it works. (Those numbers are taken out of thin air but should give you an idea why we don't recommand the "tweaking" of it.)

Koepi

[serbersan]

Originally posted by marnum
Thanks for your answers.
I encoded "Road Trip" for
1 CD (av. bitrate at about 990 kb/s) at 544x? (16:9 format)
using Xvid (latest Koepi version) and DivX Pro to compare them.

Compressibiliy Check = 2,25

AVS-file settings used for both:

mpeg2dec3
lumoff=-2
Convolution3d(preset=lowqu) Preset for noisy DVD source
LanczosResize
unfilter(5,5)

Encoding settings Xvid:

min quantizers: 2 both
max I-frame: 6
max P-frame: 16
H361 both passes
no lumi-masking
everything else = default settings

2nd pass ext. calculated by GKnot using f-corr and motion-corr.

Encoding settings Divx 5.02 Pro:

min quantizer: 2
max quantizer: 10
GMC & Bidirectional used, no Qupel
PVE = light
RC averaging period = 7500
everything else = default settings

Result:
Xvid looks worse than Divx, less detail and more artefacts. Divx simply looks better, when I added noise it looked just like the DVD (or even better, because the DVD had some blackblocking sometimes). Maybe I've left out some vital settings, but I was quite disappointed by Xvid. Divx is easier to use and there ain't so many settings where you can do so much wrong.


I've done Road Trip too. But with Xvid from Nic 21/10 build only with qpel, no bframes. The resolution I've used is bigger.

XviD: qpel, no bframes, linear scaling no lumi masking.
Divx: qpel and bframes only, no gmc or other features.

My results:
XviD was better and have retained more detail than divx. I've used the awesome modulated feature. Better quality in XviD than only using h263 or mpeg in both passes.

The comparation between images show a little more detail with Xvid, but playing the movie was more pleasure with XviD clip than with DivX... not a little difference. Well I think that when b-frames are fully implemented in XviD Divx will be completely out of the game in quality metrics vs XviD

But again: this was using qpel in XviD no an stable build

I think there is no reason to use lumoff in this movie or convolution, I've used C3d("preset=hq") and without it the quality is better. But personally I'm not in favour of the use of filter even if them are such as good as c3d.


:devil: XviD it's an evil¡¡ woow¡¡:devil: :eek:

[marnum]

>>Divx: qpel and bframes only, no gmc or other features.

You don't have to choose Qpel, but you should use GMC (ask some guys at divx.com)

>>you have all hints in post 2 and 3 in this thread - why another guide? iago mentions that luma offset stuff in his "quick start 2pass guide", so it IS already covered.

So why not combine 2nd and 3rd post of this thread with iago's guide?

[yaz]

Originally posted by marnum
@ yaz:

I don't need a wise ass telling me that I've done everything wrong. Be happy god has blessed you with so much knowledge.


i'm not the wise ass telling u that u've done everything wrong, because i'm not wise, i'm not ass (however i've got one:-) & u've done only some things wrong. be happy God's blessed u with so much self-esteem & self-confidence.
anyway, i haven't noticed that u'd answered any of the qustions put on u.

y

ps i'm just translating a book on the history of english criminal law system. the word 'offend' is used routinely for criminals. (just occured to me)

[iago]

@marnum,

First, starting with the positive part of my reply, in the next version (update2) of the document I may include some information on editing the avisynth scripts (which are mostly produced with GKnot I guess) and using Marc's MPEG2Dec3.dll with some sample parameters. I know it is a PITA for especially new users to deal with avs scripts in the beginning. Recorded in the to-do list ;).

Second, the negative part of my reply: yaz is undoubtedly a "wise" ;) and respected member of the XviD and the good old Nandub community; however, as he has already replied in his post, he is not an "ass", though he -like all of us- has one certainly! LOL ;) He's only trying to help others here, and that's something to be respected not something to be criticized imho. So, please watch your language when you address other people around here, and choose better words from your lexicon when you write to the forum.

regards,
iago

[Nic]

Dont worry, Koepi already used the strike button in that situation. I thought yaz was actually being quite helpful & informative. But lets keep on topic & leave that alone now otherwise ill have to close the thread.

Cheers,
-Nic

[iago]

@Nic

Yes, I think you're right, and it's time we went back to the topic of the thread. As a result...

----------------------------------------------------------------------
To optimize our encodes for TV output and to get rid of the black-blocking problem (which is much more visible on TV) as much as possible:

**Assuming most users use GordianKnot to generate their avs scripts**

1. First download "mpeg2dec3 beta6.zip" from Marc FD's site: http://ziquash.chez.tiscali.fr/ and then unzip the MPEG2Dec3.dll file in the package to your GordianKnot directory (where the original mpeg2dec.dll that comes with GordianKnot exists).

2. Open your avs script with notepad and delete all the lines that start with # symbol. The cleaned-up avs script should look something like this (if you don't use any denoising filters, etc. - and the paths as well as crop and resize values will/might be different of course):

LoadPlugin("C:\PROGRA~1\GORDIA~1\mpeg2dec.dll")
mpeg2source("D:\MOVIE\MOVIE.d2v")
crop(4,8,712,464)
BicubicResize(640,352,0,0.5)

3. If you have a clean source, modify the first two lines of your avs script as below:

LoadPlugin("C:\PROGRA~1\GORDIA~1\mpeg2dec3.dll")
mpeg2source("D:\MOVIE\MOVIE.d2v",lumoff=-2)
crop(4,8,712,464)
BicubicResize(640,352,0,0.5)


4. If your source contains some blocking artifacts etc., you can modify your script as below (to apply some pre-processing as well):

LoadPlugin("C:\PROGRA~1\GORDIA~1\mpeg2dec3.dll")
mpeg2source("D:\MOVIE\MOVIE.d2v",cpu=4,lumoff=-2)
crop(4,8,712,464)
BicubicResize(640,352,0,0.5)

5. Save your modified avs script. Then load your modified avs script in VirtualDub to start encoding. From this point on, you can follow the "XVID TWO PASS ENCODING" document.
----------------------------------------------------------------------

best regards,
iago

[MaTTeR]

Maybe the first step should be for them to ensure AviSynth 2.06 is installed? A nice newbie like installer is provided now too;)

Edit- Also, even for clean sources I always use at least something like TemporalSoften(1,3,0). This would help compressibility slightly but maybe it's too much for the newbie guide?

[marnum]

@ yaz:

LOL :sly: I've taken the wrong words - Sorry if I offended (not meaning you're a criminal :p ) you!

@ iago:

So unfilter(5,5) won't help if I'm already using lumoff=-2?

PS: I'll encode Road Trip one more time - with the guide :)

[iago]

@marnum

UnFilter (with either (-,-) or (+,+) values depending on the source) has always given me even better results (than lumoff-2 alone) when used together with lumoff=-2 parameter.

Also you can post your avs script for a final check ;) and also point out which XviD build you will use.

iago

[marnum]

Hm, I've looked at the GK compressibility check once more and it said 50 % or so for the chosen resolution (512x?), so I've stopped encoding Road Trip (It's my DVD, so I don't need it as Xvid) and at the moment I'm encoding Gladiator (for 2 CDs) which has a very good compressibility.

640x272
0.269 B/P*f = 81 % (C-check with SimpleResize, no denoising)
1171 Kb/s

I didn't use Convolution3D for C-check, but I'm using it for encoding (I guess that will increase Compressibility to ~85 %.)

AVS-Script:

_____________________________________________________________________

LoadPlugin("D:\...\Mpeg2Dec3.dll")
LoadPlugin("D:\...\convolution3d.dll")

LoadPlugin("D:\...\SimpleResize.dll")

mpeg2source("D:\...\gladiator.d2v", CPU=4, lumoff=-2)

crop(4,76,712,432)

SimpleResize(640,272)

Convolution3d(preset="movieHQ")
_____________________________________________________________________

As you can see I'm using no unfilter (Read your post after 2 hours encoding, so I didn't want to abort)

I don't know exactly what Xvid build I'm using :confused: , but I think it's the latest stable Koepi's build. It has no b-frames ability...

Xvid settings:
Exactly as described in Iago's guide

Let's see what Xvid makes out of it (~7 hours remaining).

PS: :scared: What are b-frames for? Is it the same as bidirectional encoding with Divx? And what will it do to my movie (once it's included in a stable build)?

[iago]

@marnum

If I were you, I wouldn't use C3D (or any other filter) with a compressibility of already 81%. But it's not a disadvantage actually, since with low settings such as (preset="movieHQ") it might improve the quality of your encode even more.

The important point is I hope you use the same avisynth script without any modification for both passes. In other words, if you don't want to ruin your encode, you should use Convolution3d for both 1st and 2nd passes, and I assume you already do so.

With a compressibility of >81%, it's almost impossible for you to get an unsatisfactory result.

iago

[marnum]

Yeah,
Encoding's finished, and it looks really good! I compared some scenes with the DVD and there's not much difference - except in the beginning, when there's this brown smoke it doesn't look well (too high quantizer I guess) but after adding mplayer-noise it's PERFECT! No need for sharpening (I guess there shouldn't be at 640x?) and actually no need for post-processing (100 % does improve things even more, of course).

And yes, I didn't change my avs-script. One time I did (some weeks ago) with Divx and the result wasn't very good (I tried encoding 1st pass without c3d and 2nd with c3d, to shorten encoding time).

[cjv]

Just read your last few posts, I'm glad you're happy with the outcome...from the looks of your script, I would imagine that the encode is kinda blurry, and a has less details than it could...here's a few suggestions:
1) using CPU=4 is not always neccesary unless the DVD has nasty compression artifacts and/or blocking...in my opinion, with Gladiator this is not the case..and you could try CPU=0
2) using SimpleResize is less sharp and precise than using something than LanczosResize or BicubicResize(0.65)..although it is faster
3) using convolution3d() _after_ resizing is risky, and sometimes can wash out any fine details that may be left.

As long as comp check is around 65-70%, the XviD codec will do a suprisingly good job of making you movie look great! Also, if you have just one bad scene, you can filter that with higher settings than the rest of your movie..thus avoiding filtering unnecessary scenes

Assuming you want to filter _only_ the bad scene (w/smoke)..try something like this (just substitute the proper frame numbers):
=============
src=mpeg2source("D:\...\gladiator.d2v", CPU=0, lumoff=-2).crop(4,76,712,432)
badScene=src.trim(0,5000).convolution3d(preset="movieLQ").SimpleResize(640,272)
movie=src.trim(5001,0).LanczosResize(640,272)
return badScene + movie
==============

Remember, these are just my opinions/suggestions...do what you want with them, but you may be suprised how sharp and amazing Gladiator can look.

cjv

[iago]

@cjv

I totally agree with you regarding c3d, the resizing method, and the cpu=4 parameter, but actually I didn't want to discuss and elaborate on these, since marnum was happy with the result. And also, I want to add to the above that I would absolutely prefer to use MPEG quantization type for a 2CD encode (of a pretty well compressible movie), though I don't know what marnum used.

regards,
iago

[marnum]

@ iago:

Could you add the things you and cjv have just said to your guide (that you'd use mpeg-qu for 2cd-rips, and that cpu=0 is good for well-compressing sources etc.)?

Yes, I used Modulated HQ and I've watched some more scenes - They're all very good except the one with the smoke - I think I'll reencode that one with quality =85 or something like that and paste the two fragments together.

PS: I won't be able to answer your next posts this week because I'm away from home.

[Koepi]

The thing about mpeg quant for 2cd and h263 for 1cd is VERY old, and everybody knows about it.

It looks like you didn't read:
- Doom9's XviD guide
- the XviD help file/XviD options explained

thus violating forum rule 1 and then asking to duplicate that info _again_ somewhere.

Koepi

[serbersan]

I've seen many scripts with bicubicresize(..,..,..,0.6X) or with 0.5 and with lanczosresize.

Is there any problem using bicubic with 0.75??

[iago]

@serbersan

There's no problem using Bicubic with 0.75. The results and compressibility are very similar to LanczosResize, and both are pretty successful imho.

iago

[Koepi]

Lanczos vs. "sharp bicubic" still produces the "more original" colours IMO. But that's a matter of taste and not always visible.

Regards,
Koepi

[iago]

@Koepi,

I agree with you, better color representation and maybe less ringing are pros of LanczosResize compared with BicubicResize 0.75. I just wanted to point out that BicubicResize 0.75 does NOT cause "disasters" in an encode when compared to Lanczos, as serbersan fears ;).

best regards,
iago


P.S.: Btw, some ringing (which seems annoying when viewing an encode on PC monitor) can provide a better-looking view on TV, giving better contrast, as yaz once mentioned in one of his posts, which seems true in my experience so far ;).

[Sharro]

Well... more and more I see the question "When will you (Iago) put that on the guide?"

I'm beggining to think that running a search for user Iago posts is a nice way of trying to learn the results of many things we could mess with under Xvid.

Iago, you could at least had 2 avs (Regular and Tv Out Optimized) scripts to your guide even though I know it's not the purpose of a newbie guide I'm sure it will help many newbies get more confident in the avs they get from gknot.

All the best.

Sharro