ffvfw settings to create KDVDs and KVCDs
First of all, if you don't have the KVCD "Notch" Matrix in ffvfw format, download it here: http://www.kvcd.net/kvcd-matrix-for-ffvfw.qmatrix
This first screenshot shows the settings to be used for making a KDVD: http://www.digitalfaq.com/archives/i...2004/02/10.jpg This second screenshot is for making a KVCD: http://www.digitalfaq.com/archives/i...2004/02/11.jpg All the following screenshots are common to both KVCD and KDVD. Please note the field in red, which I left empty, so you fill in the correct value from the previous two screenshots, depending if you want to do a KVCD or a KDVD: http://www.digitalfaq.com/archives/i...2004/02/12.jpg http://www.digitalfaq.com/archives/i...2004/02/13.jpg http://www.digitalfaq.com/archives/i...2004/02/14.jpg http://www.digitalfaq.com/archives/i...2004/02/15.jpg http://www.digitalfaq.com/archives/i...2004/02/16.jpg The following optional screenshots, are to insert noise into the encoding. Use it to enhance visual quality, if the material requires it: http://www.digitalfaq.com/archives/i...2004/02/17.jpg http://www.digitalfaq.com/archives/i...2004/02/18.jpg To do file prediction, you change the values of "Min quantizer" and Max quantizer". The higher the value (up to 31), the lower the file size. Enjoy, -kwag |
Hi Kwag,
Just one question about GOP. Since I live in PAL I should use GOP size 15 for KDVD and GOP size 24 for KVCD, right? So, should I change my settings from "Minimux I Frame Interval 18" to 15 or I could just try it with 18 since most standalones will still play it? I mean, does it affect your KB/S calculations if I change it from 18 to 15? Thanks, man. I don't know what we would do without you around :) |
Quote:
Quote:
Quote:
Quote:
-kwag |
@Kwag
How did you discover that bitrate values are right :?: I mean, they are like completly from outer space :lol: why not few bits less or more ? And secondary, did you try postprocessing mode in ffdshow ? It produces a litle blured image, but it cuts file size of about 10% |
Kwag:
you advise to change the values of "Min quantizer" and Max quantizer" to adjust the final file size: the bigger the number, the smaller the file. Do we increase both numbers at the same time (2:2 change to 3:3 and then 4:4 etc)?. Or first the Min quantizer?. And, do we have to change I, P and B Min/Max quantizers at the same time, or first I,...?. Sorry, maybe too obvious questions. ... and, YES, I completely agree with rds_correia... YOU THE MAN :lol: :wink: |
:oops: sorry, too late.
I read your answer in the loong thread, to change ALL quantizers at the same time. Isn't it?. Make me know if understood you wrong. |
Quote:
I was playing with the bitrate value, and I set it down to 1 by error, and when I started to encode, Vdub exploded. So I did a test with a value of about 32, and encoded a clip. Then I increased to 10000 and did another clip. After comparing the differences in bitrate viewer, I noticed the difference in VBV. So I did a simple math to scale the wanted VBV size, and BINGO, right on target :lol: Quote:
-kwag |
Quote:
Change all values at the same time ;) -kwag |
Quote:
I know that this is from parallel thread, but IMO it's related to this thread. |
Quote:
The "Standard" value is 40KB, which diplays "20" in Bitrate Viewer. So first try the values I posted, and tweak from 20 to 24 ( 40KB to 48KB ). -kwag |
... sorry me if being too "asking", but I'm so excited with so fast improvements in this nice codec... anyway, I think, looking at the "visits" numbers of these threads, I'm not the only one excited around here :lol: :lol: .
My question: any recommendation about scripting?. Are we using MA script for KVCD?, also for KDVD, or just resize?. In the other thread we talked about using blockbuster instead ffdshow noise. Incredible recommended (for KDVD I think) this parameters: Code:
Blockbuster(method="noise",detail_min=1,detail_max=20,variance=1.0,seed=4326) Maybe it's still to early to talk about scripting, when we are still trying to adjust the codec. |
Quote:
|
For 24 set 1383 in bitrate. Do you agree Kwag? This is what BV displays. Is this affect a lot comlpiance with standalones ?
|
Quote:
I did all my KDVD till now with GOP=25 and all my three players play it very well. |
Quote:
So if you want a value of 48KB, do this: (1147 / 20) * 24 = 1376.4 So round it to 1376 and that will give you a new VBV buffer size of 48KB (displayed as 24 in BitRate Viewer) -kwag |
Wow, great kwag!
One question. If we change min. and max. quantization parameters at the same time, wouldn't that mean that each scene (slow and action) would be encoded with the exact same quantisizer? :roll: Is that something we want? wouldn't it be way cooler to encode with a variable quantisizer? Just a thought... btw: your samples actually speak for themselves. just curious though! :mrgreen: |
Quote:
Quote:
-kwag |
I definitly have a problem with this codec :!:
All settings set as on the snapshots above (except for max quantizer in macrobloc that I can't change and is set to 31 and not 2 like on the snap). :arrow: My Q curve it like a saw blade ! Values vary constantly from 0 to 2. The picture is very bad compared to the same sample made with CCE (and the file size is exactly the same) : tiny mosquito noise around edges and lot blocks in dark scenes (even with noise in ffvfw or in the script). BTW, the sample was 53sec for a 90 min movie that is 1% sample. The file size is 16 Mb. So the final movie will be 1.6 GB without audio. ONLY TWO 90 MIN MOVIE ON THE DVD ! I drop this codec since the "Q curve" shape is not solved. If someone have an idea... |
Hi Phil,
Go to "Masking" option, and select "Luminance masking" and "Normalize adaptive quantization". Now go back to "Quantization", and you can set Macroblocks to 2. Then go back and deselect "Masking" options ;) -kwag |
Will this affect filesize or quality??
Anerboda |
Quote:
Still more options to play with. :) But with the quality and space we're currently getting, I'm a VERY happy camper :drink: :lol: -kwag |
Alright it's 22 minutes after midnight here, but after the sample I have just encoded I'm gonna have a goodnight beer. 8)
Edit: I see I don't have anymore in the freezer. Gotta take one out of the fridge then. :mrgreen: (just joking, i'm not an alcoholic) |
Quote:
Meanwhile I checked the sample you post in the top of this thread. It is actually as bad as mine ! :-( One more time am I the only one that see the mosquitos (arround the title at the begining) ? The blocks (on the wall just above the same title, on the wall above Val Kilmer shoulder at sec #6, on the blue wall on the right of Carrie Ann Moss sec #8 and the most awfull, the brownish area on the right, under the "thing" where AMEE is writen, sec #9) ? The ants on the faces (Carrie Ann Moss, very last second of the sample). I was fearing my sample was to complicate but I see that it is the codec that is not so powerfull. (note : I use the same sample for all my testings since I started KVCD). EDIT: Okay, my quantizer is set to 2/2 also for macroblocks now. But same result : the Q curve is a real saw blade. Average just dropped from 1.05 to 0.96 but max is still 2 and min 0. I never had a linear (or near linear) Q curve with this codec, whatever the sample I encode (PAL and NTSC). |
Hi Phil,
I want you to try something. Because the DCT precision in this CODEC is only 8 bits, change ALL values on the INTER portion (right side) of the matrix to "16" (the standard values for MPEG inter values). So now the "KVCD Notch Matrix" will work only on the INTRA part. Your encoding speed will now be lower (more frequencies are now being encoded), but the quality and MAX bitrate should be higher. Tell me your results :!: -kwag |
Quote:
I guess is the alarm that tells me "go to bed" :-) BTW, I have to remove what I said just above : I set all values to 3, the sample size drop to 11 (insteed of 17) and with this size, CCE is... hum... I didn't even try to reach that value with it actually :-) See ya tomorrow. |
Quote:
But Phil, ... look at these samples: http://www.kvcd.net/forum/viewtopic....=asc&start=285 I first thought only on high CBR bitrate an high min quantizers the Curve will be constant, but my last one cd encoding sample on "The Core" showed thats a "settings" thing and I gonna find out how. Also read there as I explained something about thats not important to archive a "constant" Q curve .... also a quote of Kika where he explains that Bitrateviewer thing. Good night man! :) |
Quote:
Quote:
Quote:
Remove all noise (in ffvfw or in the script), go in Motion Estimation tab and check all "Chroma" box. Change the shape of EPZS and ME diamond to "default". And a little gift : the encoding speed raises while the file size decreases :-) I'll do more test tomorrow. Bye. |
Kwag pal..
Very interesting thread.. 8) Quote:
:idea: :idea: Can you sticky this on the first page of THIS thread (and the other one, which seems to be related) on that one's first page also ?? . . I had a hard time figuring out how to mux (still haven't done the mux yet) but now that I do, I can't wait to try it :P . . I've ben messing w/ this (finally got it figure out w/ MPEG-2 encoding) no thanks to anyone here :( ..but.. .. .. I'm getting there anyways. Yeah, I've ben doing my Star Wars AtackClones (Yoda scene :P ) and I'm making a bit of progress. I followed your first steps (on page 1) and then after reading Dialhot's, I got it down minus 2mb - - so that was a good step in the right direction (I think) congrats to you and Dial and others here and other thread on this ffVFW toy :P -vhelp |
- - Request - -
can you make a quicky Matrix .txt file (or whatever fomat ffVFW requres) (left pan and right pan (both)) and post it on first page of thread so that we can D/L it, instead of having to type it all in ?? ?? I'm still using the defualt matrix, and I'm already going blind. Thanks. -vhelp |
Quote:
-kwag |
Quote:
Now you have A LOT of reading to catch up ;) -kwag |
Reading, you say ??
..hehe, that's all I've ben doing, but I did skip a few pages. (much faster when I'm surfing at work :P ) Listen, I did some testing, using MPEG2VCR. When I first try to open my .mpg (avi wrapper) file in Bitrateviewer (or TMPG's tools) they won't open for MUXing. I decided (since I don't have DVDpatcher) to try MPEG2VCR for a test run. The clip muxed pretty fast, but there seems to be some hesitation. I'm sure that my next step would be to do a real MUX job. But, why do I need to do a 2nd MUX 8O - so, I'll try and find a good muxing app, perhaps bbMPEG. Listen, if you're using bbMPEG, are you using the DVD settings, or MPEG-2 settings ?? Wait.. Are you first muxing out your video/audio and then muxing them back together, and then DVDpatching it, and Pulldown'ing it. hmm... Maybe you should cook up a A-Z receipee, so us dummy's can follow in as shorter time 8O Thanks for putting up w/ me :P -vhelp |
Quote:
i've just had a look at the help file and there is an info about Chroma: Code:
Warning: chroma doesn't work with b frames currently EDIT: @All: Have you already tried SATD with a higher diamond :?: |
Quote:
But I've used SATD with adaptive size 2 and 3. Dont know what will be with others diamonds. To Phil and anybody intrested With all this settings try MPEG1 instead of 2. 50% slower, but results ... 8) |
Quote:
|
I know :)
But for KVCD (as in the title also), you always say that MPEG1 is beter. This is my point. |
Quote:
|
Quote:
-kwag |
Quote:
-kwag |
With (S)VCD resolution and the same file size, I bet on MPEG1, even without notch matrixes. I've made MPEG2 test with settings you propose, and they're great, almost pure. But try MPEG1 on small resolutions and you'll se difference. I don't se this "color stairs" effect (with MPEG2 also hardly noticable), and I don't use any noise filter, only ffdshow postprocessed source. Plus there is less mosquito noise.
As always - my eyes are subjective 8) |
Site design, images and content © 2002-2024 The Digital FAQ, www.digitalFAQ.com
Forum Software by vBulletin · Copyright © 2024 Jelsoft Enterprises Ltd.