New Quantization Matrices - EQM V3 series (last update: 04/02/2005 - V3ULR) [Archive] - Page 5

g8o4lyf

9th November 2005, 19:12

Hello again. I just tried out the ULR matrix to rip Batman Begins to a half dvd5 xvid rip and I must say I am very impressed. I was planning on using the HR or UHR matrix but I noticed the rips being softer by those matrices. I went through this post and I saw the screenshot posted by crashuk and the sample looked very sharp to me so I gave ULR a shot. To make things harder I ripped the dvd at 816x336 anamorphic@1590kbps (thats the bitrate vdub shows me in properties). When I ripped the movie at higher bitrate using HR last time I noticed many many artifacts mainly because of the bats. This time around the rip was sweet. No artifacts noticable by me so far. If anyone wants any samples let me know I will post some. But I think I will stick to the ULR if it keeps producing sharp results like this. Thanks sharktooh :D

Backflip

14th November 2005, 13:01

What would you recommend for backing up Star Wars Episode III: Return of the Sith to 2CD's, Sharktooth? I'm keeping the anamorphic frame.

Numbers I'm getting using ULR V3. 2 b-frames and AQ enabled.
Source: Retail R4 DVD @ 25 fps
Average bitrate - 1009 Kbps
0.138, 66.7 of 0.206
Lanczos Resize

After cropping mattes including lighter pixels/mess (74,2,74,0), res of 704x416 is possible without going over the 100% H-Zoom.

I have a feeling it'll look alright since it's a pretty dark movie and there's CG all over the show.

BTW, Sharktooth, is it correct to crop away mattes as much as possible, including the messy bits, and then if this results in a 101% H-zoom, go down to the next 16x16 compliant resolution? i.e. 704x416.

Thanks :)

Backflip

15th November 2005, 14:02

Very surprising, the quality is splendid for a 2 CD rip. Having a high quality source must've helped quite a bit too.

Really magnificent stuff Sharktooth :)

Now that I've done a few encodes and stuff I really don't get the reall advantage of square pixel at all. I think anamorphic is the way to go + a high quality CQM.

(One sad thing that I saw which is unrelated pretty much and that's XBMC scales the picture to a 2.36:1 according to the onscreen info, on a FS PAL 25 inch TV. Don't ask me why it does this because MPlayer on the PC scales it to 980 x 416 @ 2.35:1. I'll ask about this prob at the XBMC main site.)

Sharktooth

15th November 2005, 14:27

Well, i find it quite weird XBMC behaves that way, but maybe XBMC devs preferred it or it's just a bug, or it's just a compensation for the TV scale...
However i agree, anamorphic rox.

Backflip

16th November 2005, 06:37

Yeah, it's a little strange. MPEG4 Modifier reports - Custom pixel shape (32:23 = 1.39130) and 2.355 : 1. Maybe 2.355 is a little too close to 2.36.

Anyways, thanks. I'm going to encode another movie - Ice Age @ about 1,950 Kbps (2CD's, 2 audio tracks). Think EQM V3UHR is the only option for this one. That or EQM V3UHR.

Cyberace

16th November 2005, 08:29

XBMC scales the picture to a 2.36:1 according to the onscreen info, on a FS PAL 25 inch TV. Don't ask me why it does this because MPlayer on the PC scales it to 980 x 416 @ 2.35:1. I'll ask about this prob at the XBMC main site.)Have you tested the same files in Media Player Classic (http://sourceforge.net/projects/guliverkli/) (as MPC scales very similar to XBMC)?
btw, have you "calibrated" the pixel-size to a perfect square in XBMC settings before playing?

PS! Understand scaling is much more correct in XBMC and Media Player Classic than in MPlayer!

Sharktooth

16th November 2005, 13:48

Backflip

20th November 2005, 05:56

Is there anything you'd recommend not using when it comes to encoding animation such as Ice Age, Sharktooth? Is Cartoon mode ok?

Sharktooth

21st November 2005, 21:55

keep cartoon mode for toons and low detail animes with large flat colored areas not CGs.
CG needs smooth matrices (EQM V2 for example).
Avoid oversharpening/overfiltering, usually CG movies are already clean. If they have some mosquito noise (usually added in production phase) just clean it with non aggressive temporal filters (temporalsoften or something).

Backflip

28th November 2005, 03:07

Ok, I'll try that with Madagascar. Figures are pretty low using UHRv3:
0.197, 41.7% of 0.471. Using AQ and 2 b-frames.

Meat_PoPsiclez

29th November 2005, 01:49

Description:
This is an EXTREMELY HIGH BITRATE matrix designed for constant quality encodings ONLY. Compressibility is "ridiculus" in comparison with other high-bitrate matrices but there is a reason...
It's application is pretty obvious: it's for backing up original material or HD material without using lossless codecs.

Well I tried eqm_v3ehr on some unprocessed minidv footage, and with xvid set to q=2.5 no b-frames (just as a starting point, haven't experimented with them with high bitrate encodes) I managed a 'humble' average of 10.5mbit/s covering slow motion, many abrubt scene changes, and lots of high motion. Did I mention it was noisy minidv footage, and interlaced? A lot to contend with. In any case, the results are damn near flawless. I'm not talking just perceptibility either, I'm talking frame by frame comparisons done in photoshop. Here are the results.

http://www.anti-logic.com/phpBB2/viewtopic.php?t=699

As you can see much of the image is literally bit exact! Hats off to you Sharktooth! Going from 29mbit/s (ntsc dv) to 10mbit/s with so little loss to call it nothing at all? I'm sure a cleaner source would provide even better results. Now if only camcorders could use mpeg4 with a custom matrix for direct encoding... pipe dream.

What would the point of diminishing returns with this matrix be, in terms of q setting or bitrate? For long term archival of footage for later editing, this seems to be a clear winner in my eyes.

Sharktooth

29th November 2005, 04:46

The purpouse of EHR was exactly this: near lossless encoding for archival purpouses.

Backflip

1st January 2006, 08:59

Sorry to bother you Sharktooth, but I've just realised something now with regards to aspect. When I put the PAR in the XviD settings should I not be using the actual AR which is displayed when PAL anamorphic (16:9) is checked? Example:

http://img501.imageshack.us/img501/4940/aspect015sp.th.jpg (http://img501.imageshack.us/my.php?image=aspect015sp.jpg)

So in XviD codec I should use 183 : 100 if I want to be as close as possible to the original AR after cropping? Sorry this only just occurred to me to ask the question.

Also, in this instance can I get even closer to the original AR by putting something like 183 : 10* or something?

One last thing, I have another movie that indicates 2.379 (even though on the box it's 2.35:1). Should 237 : 100 be used or 238 : 100? I personally think 238 : 100.

JnZ

1st January 2006, 18:36

Thx Sharktooth for matrices. I use them very often ULR,LR,HR,UHR for DVD's and now for HD's. Before this, I used Bulletproofs HQ matrix, but now, my encodes are litlle more scalable.

Thanks very much. Highly appreciated.

IMHO I think, that this is best matrices I've ever used, and that I encode to xvid couple of years (since B frames discovered :) and SBC was overcomed).

Sharktooth

1st January 2006, 19:00

Backflip

2nd January 2006, 06:20

How do you find out the AR of the VOB files? Thanks :)

Backflip

7th January 2006, 15:57

I think I've figured it now. Appleseed for example. I crop mattes and end with the highest resolution of 704x560. I then use a PAR of 16:9 PAL (my source is PAL). Playing in MPlayer I get:

Movie-Aspect is 1.83:1 - prescaling to correct movie aspect.
VO: [directx] 704x560 => 1024x560 Planar YV12

(1.832 is indicated in GKnot if PAL anamorphic 16:9 is selected)

It looks to be correct, is it correct? I think it is, hopefully it is :)

Sharktooth

7th January 2006, 22:35

it is.

CrashUK

4th February 2006, 14:09

Been doing some testing using your EQM V3ULR (revision 3) matrix with KDVD profile but with your. it gives lower file size than kdvd matrix. picture looks same to me.

Blue_MiSfit

6th February 2006, 21:21

Actually a camcorder that recorded directly to MPEG-4 is a pretty bad idea. Because you have MPEG GOPs, (I, P and B frames), editing becomes difficult. With DV, frames are only spatially predicted, not temporally (all I frames). If you have P and B frames, editing becomes much more difficult because you cannot directly seek to (and cut) any frame without re-encoding. It is possible (see HDV - a high def format that uses long GOP MPEG-2) - but there is some re-encoding involved, which can result in a lot of quality loss through subsequent generations.

A good workaround is to convert to a lossless or quasi-lossless format like Apple Intermediary or Huffyuv, do the editing, and then re-encode, but at this point I'm really getting off topic :) Sorry, I'm in a Final Cut Pro class ATM. A good thought though.

-Misfit

Well I tried eqm_v3ehr on some unprocessed minidv footage, and with xvid set to q=2.5 no b-frames (just as a starting point, haven't experimented with them with high bitrate encodes) I managed a 'humble' average of 10.5mbit/s covering slow motion, many abrubt scene changes, and lots of high motion. Did I mention it was noisy minidv footage, and interlaced? A lot to contend with. In any case, the results are damn near flawless. I'm not talking just perceptibility either, I'm talking frame by frame comparisons done in photoshop. Here are the results.

http://www.anti-logic.com/phpBB2/viewtopic.php?t=699

As you can see much of the image is literally bit exact! Hats off to you Sharktooth! Going from 29mbit/s (ntsc dv) to 10mbit/s with so little loss to call it nothing at all? I'm sure a cleaner source would provide even better results. Now if only camcorders could use mpeg4 with a custom matrix for direct encoding... pipe dream.

What would the point of diminishing returns with this matrix be, in terms of q setting or bitrate? For long term archival of footage for later editing, this seems to be a clear winner in my eyes.

Terka

28th February 2006, 10:14

1.) Which matrix should i use for DV->720x576 Xvid interlaced?
I want to keep the quality.
2) which quantitizer should i use for encodings? I will store on HDD
Help please.

Audionut

28th February 2006, 10:27

1.) Which matrix should i use for DV->720x576 Xvid interlaced?
I want to keep the quality.
2) which quantitizer should i use for encodings? I will store on HDD
Help please.

1. depends on excatly what size restrictions you have.
2. "2"

Terka

28th February 2006, 11:24

i want to keep it interlaced, dont want to resize and quality similar to original, sharp picture. where the matrices for interlaced encodings can be found? are they same for xvid and mpg2 encoding?

Audionut

28th February 2006, 11:43

http://rapidshare.de/files/14332073/Audionut_s_max_16_interlaced.xcm.html

Here is an high quality matrix designed for interlaced encoding.

Keep in mind that it will eat space.

Your best bet is to select a sample of your movie that represents the entire movie and encode that at quant 2.
If the resulting bitrate is to high, then choose quant 3. etc, etc.

JnZ

28th February 2006, 14:43

1.) Which matrix should i use for DV->720x576 Xvid interlaced?
I want to keep the quality.
2) which quantitizer should i use for encodings? I will store on HDD
Help please.
Hi, my countryman! :)
1) Why do you want to store it as interlaced? I think good deinterlace method done better job. Sharktooth HR,EHR are very good.
2) Use Q2 if you don't care size, but it eats lot of space.

Graal_CPM

3rd March 2006, 21:20

:goodpost: This eqm_v3ulr_rev3 matrix rocks!!!

I have almost managed to reach my goal of having a simple "universal", brainless, quick way to backup my DVDs...

And with the following avisyhtn script, I have managed to gain 10% compressibility (compared to a simple Lanczos4Resize+Undot) , keeping a very pleasing and sharp picture :

#Crop/Resize
crop(0,74,718,428,true)
Lanczos4Resize(704,288)
#Denoise
RemoveGrain(mode=1)
LRemoveDust_YV12(17,1)
#Sharpen
dull=last
sharp=last.ModerateSharpen(0.3)
soothe(sharp,dull)
#SpatioTemp smooth
FluxSmoothST(2,2)

I have used recommended and default parameters here... Let me know if I can do better. Again, I know there is no such thing as a "universal" script, but my goal here is not full quality archiving but preserving my DVD from kids-fingers, being able to enjoy the movie, and not spending too much time on each DVD...

Note that encoding time can goes up to 6:30 hour for a 133min movie on my P4-2.4 with GKnot, but it's worth it... Credits also goes to Didée for his awesome scripts of course!!

Teegedeck

3rd March 2006, 21:50

I...think you're denoising too much. Definitely. Much too much. With that ULR matrix you should be able to just keep the full anamorphic resolution without resizers as well. Should look better without such a long script.

Sharktooth

3rd March 2006, 22:35

ULR tends to artificially preserve sharpeness even if someone could say that effect is due to "ringing", but i like the result.
if you want a smoother picture use jawor's 1cd matrix.

Didée

4th March 2006, 00:22

Well, there's not that much denoising happening through that script ... surely it's not "invisible" denoising, but it's still rather moderate, IMO.

The key question is: How strong are you compressing, Graal_CPM? Are you doing 1CD encodes, 2/3/4 movies per DVD, constant quant encoding for storage on HD, ... ??

Graal_CPM

4th March 2006, 17:14

Thank you for all the fine ideas here.

My aim is strong compression/low bitrate to fit one movie on 1CD for 100min movies, and on 2CDs for larger ones. The 100min "rule" depends heavily on the kind of movie of course.
Xvid is two-pass with default parameters provided on Doom9 guide, plus ULR. Using GKnot test, I try to stay around a 70% compressibility. Final result is watched through a Sanyo Z3 (VGA in) on a 3.5m diagonal white wall (ffdshow, no filtering). My eyes are not used to HDTV, so the reference quality is the one of the original DVD.

Teegedeck:
I...think you're denoising too much. Definitely. Much too much. With that ULR matrix you should be able to just keep the full anamorphic resolution without resizers as well.
I was just wondering if it was too much too... Note that I've tried to stick with the most conservative settings. I am not so sure about FluxSmoothST... It seems like a "logical" step for high compression (+3% compressibility). I welcome any suggestion on this or ideas for another more "transparent" spacio-temp filter.
Thanks for the anamorphic tip. I should have given my objectives before (Didée's post). Keeping the anamorphic just "costs" to much for my goal. I keep that in mind for other projects though.

Sharktooth:
ULR tends to artificially preserve sharpeness even if someone could say that effect is due to "ringing", but i like the result.
if you want a smoother picture use jawor's 1cd matrix.

I think ULR result just rocks !! :thanks: Mostly IMHO because the increased compressibility allows for a larger resolution. And on all my encoding tests, the larger the resolution, the better the sharpeness... Filtering appears as a very good complement to ULR in order to kill the blocking with a higher bitrate. It is true also that without filtering (almost, just undot), I see a triffle more "ringing" when zooming (I really have to look for it on specific frames, like poles on a blue sky).
Now I wan't to give a try to jawor's 1cd, to see if I can drop some denoising, without too much smoothing.

Didée:
Well, there's not that much denoising happening through that script ... surely it's not "invisible" denoising, but it's still rather moderate, IMO.

Actually, your comment meet my eyes (subjective) impressions. I guess that "invisible" denoising is the holy grail of scripting! However with your heavy work, I feel you are getting closer and closer :D. It seems that Soothe helps a bit here to compensate for previous denoising. Without increasing compressibility and it's a first to me. A little aSharp or LimitedSharpen was still too much.

Thanx again for these precious input guys! :cool:

Terka

6th March 2006, 11:35

2 Audionut:
you mean interlaced encoding eats space, or this matrix?
which bitrate, quant should i use?
still tried 7000, or Q4 (all DV source)

Audionut

6th March 2006, 14:12

2 Audionut:
you mean interlaced encoding eats space, or this matrix?
which bitrate, quant should i use?
still tried 7000, or Q4 (all DV source)

The matrix will require a large bitrate.
Note that interlaced encoding requires more bitrate than progressive, also.

Q4 encoding should be fine.

The best thing to do, would be to encode a sample of the video at Q4. If you like the quality, and the final size is ok for you, then your set.
If the quality is not high enough for you, then try Q3.
Or if the final video size is to large, then try Q5.

edit: I also recommend that with a high bitrate matrix, you set the B-VOP "Quantizer offset" to "0.00"

BigDid

9th March 2006, 21:25

ULR tends to artificially preserve sharpeness even if someone could say that effect is due to "ringing", but i like the result.
I like the sharpness, but not the ringing :rolleyes:
if you want a smoother picture use jawor's 1cd matrix.
Just wanted to share my last tests, I use AGK and when it switch from sharp matrix (EQM V2) to soft matrix, it uses V3ULR.
I'm very satisfied with it as long as the bitrate (above 900kbps) and/or the resolution (above 608 for 16/9) does not go down. In that case the ringing is very present and I now use H263 instead and the ratio sharpness/ringing is better to my eyes with H263.

I had this idea re-reading the CQM tests made by soulhunter and seeing that H263 was near the V3ULR I wanted to give it a try and have really good results.

How-to in AGK: in hidden option (Ctrl-F9) use the "enable stand-alone support" and select "ESS"; ESS uses H263 (or mpeg custom) for compatibility with the SAP having an ESS chip.

Did

Teegedeck

9th March 2006, 21:35

H.263's natural predator (competitor) isn't ULR but EQM v3LR or HR. If H.263 does look better on your source than ULR, then I think you have been using the wrong matrix all along or, put differenty, it could well be that you have been using ULR for too high bitrates. You could check out whether EQM v3LR or even HR beat H.263 on your encoding setup.

Sharktooth

9th March 2006, 22:21

Probably... but he's right on one thing though: the resolution.
ULR is made to be used at higher resolutions than 608*xxx. Lower resolutions will show bigger artifacts that would be hidden in higher res.
In the ULR case the ringing will be more evident, but it's "needed" to keep the image sharpness at higher res.

BigDid

9th March 2006, 23:12

Thanks to both for your valuable comments.

It's true I have not compared H263 to Jawor 1cd, will try to (lazy me only using AGK). Also I'm quite sure soulhunter's comparison was for 800-900kbp/s (or low bitrate ), that V3ULR was winner, and H263 was just behind it; just cannot find the threads right now, will edit if/when research succesfull :angry:

Too high bitrate is not really an option for SAP compatibility (usually around 1200kbp/s and/or 640 wide for 16/9). AGK uses EQMV2 as sharp matrix whenever possible*, so in these cases I will try to compare EQMV2 to v3LR or HR.

SAP compatibility is also my concern concerning the xvid presets thingie Teegedeck is starting. Give me time to ingest that highly interesting but though thread and I will post there :)

Did

* From recalling, the AGK comptest is something like:
- If using sharpmatrix, will a 608 wide (for 16/9) be possible with a 60% minimum compressibility?
if yes use EQMV2 with sharp resizing (lanczos)
- if not will 608 wide be possible with soft resizing (bicubic?)
if yes use EQMV2 with soft resizing
- If not will 608 wide be possible with V3ULR with sharp or soft resizing for 60% compressibility
If yes use V3ULR with 608 wide with sharp or soft resizing
- if not use V3ULR with less than 608 with sharp or soft resizing and decrease resolution until 55 to 60% compressibility is reached

Edit1: gotcha
V3-900kbp/s xvid 1.02; 1st v3ulr, 2nd h263, 3rd jawor 1cd here:
http://forum.doom9.org/showthread.php?p=570152#post570152
Also V4-1100kbp/s xvid 1.03 here: http://forum.doom9.org/showthread.php?p=648109#post648109
Edit2:
The defaut "soft" cqm/matrix used by AGK is not V3ULR but Jawor1cd; it is however possible to use the v3ULR as soft matrix by creating a ".newsoftmatrix" file in the AGK install directory.

Backflip

11th March 2006, 07:39

Interesting, I used EQM_v3UHR_rev2 for The Skeleton Key recently along with qpel. Main reason for using it was compessibility was huge on the movie and I was making a 2CD rip at 672x288 resolution (yip, square pixel). Although I haven't done a comparison with other ways of encoding the movie I think it did a good job.

I'm thinking of using Autogk sharp matrix for a movie. Should I be considering using this at this point in time or just stick with the other matrix's?

edit - nvm, I've just looked at the matrix's and eqm_v3hr is pretty similar to (or might be the same as) autogk sharp.

ps - I did want to stick with full anamorphic because I do see a benefit, as far as encoding 1.85:1 movies, but I've decided to go back to square pixel. I haven't fully grasped how to encode a movie with accurate AR (which is the initial reason I chose to begin encoding in anamorphic) if it's an approx. 2.35:1 or 2.40:1 movie since you can't just put in 16:9 PAL (I think), you need to use DAR settings. If anyone could tell me where you get the DAR setting from and how to most accurately put it in that would be nice though. Is it not just the PAL anamorphic 16:9 Aspect Ratio number you get, so if it's 2.331 or 2.44 I'd put in the DAR settings: 233 : 100 or 244 : 100 respectively?

Teegedeck

11th March 2006, 10:23

Simply use (PAL or NTSC) 16:9 Pixel Aspect Ratio. That'll be enough.

BigDid

11th March 2006, 19:09

...I'm thinking of using Autogk sharp matrix for a movie. Should I be considering using this at this point in time or just stick with the other matrix's?

edit - nvm, I've just looked at the matrix's and eqm_v3hr is pretty similar to (or might be the same as) autogk sharp...
It seems we have similar concerns: you use a high details (less compressible) matrix and want to try a more compressible one. I use a more compressible matrix and would like to try one keeping more details: LR or HR

Keep in mind that AGK is aimed for low to medium birates and that most users have SAP compatibility in mind. I, usually encode between 900 to 1200 kbp/s, sometimes below (A capture for aviator was 850kbp/s and H263 :D ). I still have to try higher bitrates, and the CQM related, in conjunction with the MTK6000 profile if I want that kind of encode to play on my SAP.

Beside digging in the main AGK thread, only Sharktooth or Len0x may recall the tests making EQMV2 the winner for the sharp AGK matrix. Hey that's a good one; Sharky: at that time were LR or HR pretenders for V2 or was it really out of scope?

Did

Sharktooth

12th March 2006, 02:50

It seems we have similar concerns: you use a high details (less compressible) matrix and want to try a more compressible one. I use a more compressible matrix and would like to try one keeping more details: LR or HR

Keep in mind that AGK is aimed for low to medium birates and that most users have SAP compatibility in mind. I, usually encode between 900 to 1200 kbp/s, sometimes below (A capture for aviator was 850kbp/s and H263 :D ). I still have to try higher bitrates, and the CQM related, in conjunction with the MTK6000 profile if I want that kind of encode to play on my SAP.

Beside digging in the main AGK thread, only Sharktooth or Len0x may recall the tests making EQMV2 the winner for the sharp AGK matrix. Hey that's a good one; Sharky: at that time were LR or HR pretenders for V2 or was it really out of scope?

Did
LR and HR were already there but v2 was preferred coz of the versatility (good for movies, cg etc.) while hr and other 5 matrices blended between HR and v2 were somewhat too sharp (or there were artifacts) for some kind of sources (mainly cg) and some of them (mainly HR) gave less compressibility.

BigDid

13th March 2006, 18:09

From a few posts back:
"It's true I have not compared H263 to Jawor 1cd, will try to (lazy me only using AGK)."

I did some quick tests from 5mn Vob part: Fusion the core, mix of quick and slow motion, some faces close-up also.

Result are:
1/ULR3 More compressible, good sharp overall look, some ringing sometimes strong, sometimes soft.
2/Jawor1cd, 5.3% less compressible than ULR3 (on that clip), good overall look, softer than ULR3, nearly no ringing.
3/H263, 14.5% less compressible than ULR3, good overall look, softer than ULR3, nearly no ringing.
So it seems Jawor1cd may be the replacement for ULR3 with lower width/bitrate, quite same results as H263 but more compressible. I'll give it a try on a full movie when possible.

To Teegedeck, you are right in terms of compressibility, H263 is more competitor with LR than ULR.

Also the AGK-V2 (see Sharktooth comments) seems really balanced between LR and HR. I'll try comparing V2 to HR more extensively when possible, goal beeing is HR giving really more details than V2?

Did

neutrogenik

26th March 2006, 12:50

Thanks a lot for the matrix grids, my vid quality are much more better with it :thanks:

fight2win

26th March 2006, 15:08

while encoding with vdub, how to use these matrices?

BigDid

28th March 2006, 04:06

while encoding with vdub, how to use these matrices?
Hi,

In Vdubmod (I don't use VD) Menu video/Compression/choose xvid (way down)/profile/quantization type, change to MPEG-custom/edit matrix/ load your custom matrix (.xcm) done.

Looking for some screenshots seen recently, will give a link if found.

Did

Ps: Gotcha: http://forum.doom9.org/showthread.php?p=805048#post805048
shots are in xvid config/quantization type/drop down box is H263 have to check MPEG-Custom to get the edit matrix box functionnal :)

NeD tHe OnE

30th March 2006, 08:21

Hey Sharktooth! It seems that I have a problem using EQM v3 ULR!

Is this preset correct for 1CD Backup of 100 mins length movie, with 640*xxx backup ?

"
slow extreme compression: VHQ=4, VHQ for b-frames, AQ, EQM v3 ULR, 1st pass @q=4,
curve-compr. H30, L15, quantizer-restrictions min. 4, max. 7"

I got an undersized file like around 450 mb when i need it to be 700'

but i got exact 700 when i changed the values a bit lyk 1st pass @ wt =1 & quant. restrictions min=1 , max=7 )

Just want to ask you whether its a right thing to do...

NeD tHe OnE

30th March 2006, 08:32

One more thing? which is according to you is better H.263 or ULR v3?

for one cd backup.

H.263 gives you soft pictures it is good @ low bitrates,

and what is the optimum bitrate to go for while using ULR?

Is it better @ real low bitrates like around 400 - 800 kbps?

foxyshadis

30th March 2006, 08:40

It says all over the forum to NOT do that. You get zero extra quality and might as well just burn the 450 to a cd. If it undersizes then you go up a matrix, which will put that extra space to good use! In this case from ULR to LR.

NeD tHe OnE

30th March 2006, 08:55

I might have backed it up if the Quality was good ...

I mean imagine a video @ 400 kbps , its very low ... I can clearly make out its artifacts ... What should I do?

when i did that I got a smooth Video arnd 1000 kbps ?

Ok .. you tell me to go for a higher preset . What exactly should i do for that to get 700 mb and best quality.

Forgot to mention ... i am using Gordian Knot

foxyshadis

30th March 2006, 09:13

"quantizer-restrictions min. 4, max. 7"

I didn't see this. Well, it makes sense now. Why even bother with two-pass if you're going to do that? It should always be set to min 2, max 31, unless you have some very special reason to set it otherwise, or you will get bad video that's often under or oversized. Try a fresh 2-pass with that and ULR.

VHQ=4 is probably overkill at this bitrate, the low bitrate artifacts chew up any minor gain it could make, and you lose a lot of speed. VHQ 1 or 2 would make more sense.