Avisynth: Some No-Good Results - Page 3

jorel · #41 11-24-2003, 05:21 PM

Quote:

Originally Posted by Jellygoose

Hi Dialhot!

I use MPEG-1, and the problem of blocks at high bitrates with CQ_VBR is the only disadvantage I see at that resolution compared to CQ.

hey Jell,
now that you show i did samples with fast and low actions...
it's true

mpeg1, 352*2xx with CQ is better for high action
and CQ_VBR is better for low action

what we do now?

thanks for the link and script!
did you change something in the script?....please,post there!

kwag · #42 11-24-2003, 08:54 PM

Quote:

Originally Posted by jorel

now that you show i did samples with fast and low actions...
it's true

mpeg1, 352*2xx with CQ is better for high action
and CQ_VBR is better for low action

what we do now?

Encode all action scenes in CQ, and low action scenes in CQ_VBR. Then merge all clips together. Then we :banghead: because of sync and file size issues ( just kidding

)

jorel · #43 11-25-2003, 07:19 AM

oh yeah!
:banghead:
i did it......now i need "band-aids"

.......

Jellygoose · #44 11-25-2003, 09:24 AM

This really sucks...

I'll do more testing...

Dialhot · #45 11-25-2003, 09:33 AM

Quote:

Originally Posted by Jellygoose

This really sucks...

I'll do more testing...

Did you try to change the motion precision search ? I guess that you should have a decent result in CQ_VBR by using High quality insteed of Motion Estimate Search.

Jellygoose · #46 11-25-2003, 09:54 AM

I do all my encodes using "High Quality"

Thanks for the tip though!

incredible · #47 11-25-2003, 11:33 AM

Did you turn "Detect Scene Change" OFF?!!!
Also try a GOP 1-12-3,maxGOP=0(unlimited!) .... as it also contains more Bi-Directional Frames which also refer to the next I Fame this could still preserve the Quality.
When you have set these options check the quality of the Predicted sample (xxx.m1v.sample). If Quality is still good try another prediction using an even higher GOP like 1-20-3 ... maybe you'll still receive a good Quality.
But using these GOPs the image will be a touch "softer" so lower the values of mergeLuma!

kwag · #48 11-25-2003, 11:36 AM

Quote:

Originally Posted by incredible

Try also a GOP 1-12-3,maxGOP=0(unlimited!) .... as it also contains more Bi-Directional Frames which also refer to the next I Fame this could still preserve the Quality.

Nope. Don't use 3 B frames. Use 2.
Read here: http://www.kvcd.net/forum/viewtopic.php?t=2073
Sorry, it's the long thread, but it's all documented there

-kwag

incredible · #49 11-25-2003, 11:40 AM

I know! And I know the thread well. But I had some experiences on some movies where it resulted ok by also obtaining more compression ... thats why I said to him "Check the Quality" ... And in here we're going to try everything so maybe Jellygoose gets nearer to his success. But maybe in case of Notch Matrix this won't give advantage. Its just a try

kwag · #50 11-25-2003, 11:49 AM

Yes, you will get more compression, but at the expense of more artifacts

That's why I pointed to that thread. Mind you, in regular TVs, there's probably no visual difference. But on large TVs, you can notice the effect.
But if you want to go for more compressibility, then go for it

-kwag

Jellygoose · #51 11-25-2003, 01:17 PM

Well thanks again everybody for chiming in, and trying to help, but I just think I'm already at the very limits, with the current abilities we have.
I don't think it is possible to get such a high action movie (length 210 min. w/o credits) on a 99 min. CD-R, with a perfect picture (of course I mean perfect for that resolution 352x288, no mosquitos, no blocks, no dancing DCT blocks...) with the current "tricks" we use...

Maybe YMPEG, when it works will make it possible, because it's all about the BAD bitrate allocation of TMPGEnc... One last idea... do you guys think MVBR will be an option?

Encoder Master · #52 11-28-2003, 03:52 PM

@JOREL

Quote:

thanks EM,
i got the sample and what i see is:

perfect aspect radio in powerdvd and zoom player!
the matizes are very cool.
what filters are used in the script

removing doubts(i always have):

if the sample have 0:59 seconds with 5,49mb(5.631kb)without audio,
how you put this all (168 minutes) with audio in only one cd?

thanks!

It's was a fast moving scene and I pick 112kbps for the audio. I used the optimal script of kwag.

Tomorrow, if you want, I post screens of Harry Potter 1 in 480x576 on one CD-R 90! With Audio 112kbps.

jorel · #53 11-28-2003, 08:26 PM

great.....waiting for you!

Encoder Master · #54 11-29-2003, 06:01 AM

Here the screens of Harry Potter 1:

http://samsonite.webspace4free.biz/harrypotter1.html

The screens are 480x576 and so they looked ecked. If you rezide it to 720x576 it's perfekt.
I used the optimal script of kwag and max bitrate of 2200kbps (I've testet and in this resolution the video doesn't need more.

)
and min bitrate at 300kbps.

jorel · #55 11-29-2003, 09:20 PM

very very cool quality EM!
the second picture(fast action) is really clear,
the general details in all pictures are good.

..let me see:
was 480x576,optimal script,min300&max2200,audio 112,cdr90!
right?

and the CQ that you got was ...

is mpeg1 or mpeg2

i will re encode again like you,
i did 352x240,audio128,CQ_VBR,cdr80!

Encoder Master · #56 11-30-2003, 03:35 AM

THX.

With 352x240 you have also a realy good VCD Quality because you can choose a high CQ.
The Movie has many dark scenes and so CQMatic has predicted a CQ of 74,xx.

Very cool. and I've put the Sample of CQ Matic in BitrateView ans see that the Max Bitrate goes not over 2100kbps only at the beginning.

Tomorrow I post my results of Harry Potter 2. But that's quite harder to kompress.

Why you don't use 352x576. I think that will going on one CD-R 80 ind very good Quality or your player doesn't support it.

Encoder Master · #57 12-01-2003, 03:55 AM

Here:

http://samsonite.webspace4free.biz/harrypotter2.html

In SVCD Quality with optimal Script and 2200kbps Max. 300 Min. 75,08 CQ
on one CD-R 90 (I can burn 950 MB on ONE-CD 90

)

kwag · #58 12-01-2003, 04:37 AM

Hi EM,

Those shots look great

Was that encoded with the current MA script, or with the older MA script

-kwag

Encoder Master · #59 12-01-2003, 06:52 AM

THX but it's not yet perfect because some are very Noise (dark scenes).
I've to test more than that to become perfekt results.

I've used the older Optimal Script of you.
But I've downloaded the new and make some test with it.
The Scripts are very cool. I have to say it.

incredible · #60 12-01-2003, 11:45 AM

The new MA compresses even better! I did a lot of tests and the results are very very nice, ... if the case of blurry static foreground objects when fast moving objects behind occur ... won't hurt you.