Ah, ok, I missed that "D" :lol:
-kwag |
Alright then.
Just finished my first tests a couple of minutes ago. All I can say is 8O 8O 8O ! Did a manual predicition to fit 2 ~120 min. movies to 1 DVD-R (easy task I know) and my desired sample size was something about 36mb. Now with ffvfw at Quality 100 Max. Quantizer 25 the file was at about 27 mb, and already looked just as good as the TMPGEnc sample I made. We'll see how accurate prediction is, and how good the final encode will look! Now again I see VERY much space for possible optimizations here! Is MPEG-1 encoding with this Codec really that bad? I'd love to see that beautiul motion estimation deal with an ULBR Template. :mrgreen: |
Damn, how do I increase the filesize? :lol:
I'm already at Quality 100 and maximum Quantizer at 8. Filesize remains the same... :roll: |
Quote:
Shorten the GOP size. -kwag |
Well the max. Quantizer somehow is not recognised... It doesn't matter if you set it to 2 or to 25. Filesize stays the same.
I can't seem to reach an Average Bitrate above ~1700kb/sec. What I need would be 2100kb/sec. This sucks... Although I gotta agree that quality is already superb... 8) |
Quote:
I think I'm going to have a look at the quantisizer source :cool: -kwag |
Geez
Wouldn't it be nice if we could chat with Milan :?: That would save us a lot of time... I've made a decision. I'm not using tmpg again at least if things stay as they are. I'm having the same quality/speed results with both encoders and I think ffvfw still has a lot of potential to explore. I don't even need tmpg to demux since I got my hands on bbtool19. Anyway somebody having jerky playback after demuxing/muxing mp2? I do sometimes and I haven't figured out why... Maybe something to do with image processing by ffdshow. I was trying to use noise and dct from ffdshow. I'll try again doing it with avisynth. Any other tool I can use instead of VdubMod? C ya |
Quote:
Try to change to "Quantizer" mode instead of Quality Mode, cause if quality = 100 that means in Quantisizer mode "only" Quantize 2! Try Quantize 1 but the Filesize could blow up! But how does the picture look at Quality 100 ? Is it not ok? Do you just want to end up with a 100% filled up DVD-R? Thats the point of that encoder ... you can't really determine the min/max bitrates or a good range of quality where the steps within a range do give good outputs for a good prediction. IF you want end up in larger filesize with "sense", try to tweak at senseful parts of the encoding and in that case a "switching" to a std. mpeg matrix would be the best choice as your result will end up even sharper! 8) @rds_correia Well chat with him? ok, but even a mail contact seems to be not possible. (I know what you meant) ;-) Same thing like in case of Yusuf's ympeg. In here people are doing the biggest cross-world betatesting and almost no effincient contact is possible with the author, .... so we only can see/figure out Bugs but for further developement the author should be in real contact wih us or other testers. :( |
Doesn't suck anymore!
Good news to all :D
After playing around with this thing for a couple of hours, I've found how to increase the bitrates, way up to DVD levels (+8,000Kbps) on peaks. These are the settings you need: http://www.digitalfaq.com/archives/i.../2004/02/6.jpg http://www.digitalfaq.com/archives/i.../2004/02/7.jpg If you notice, I've turned OFF "Custom quantization tables", because KVCD's notch matrix just cuts off too many frequencies with this encoder. I'll revise that some other day, when I get to talk with Milan. In the mean time, even though it says "one pass - constant bitrate", IT'S NOT :!: It's actually VBR, and the VBV buffer settings are correctly set with this mode 8O ( a "feature :lol: ) Now to do some kind of prediction, just change the "Max quantizer difference", which I have set to a value of 10. The lower the number, the larger the file size. And vice versa. Have fun :D -kwag |
The end of the other encoders.
Well, maybe not for everyone, but for me, it's a big YES :mrgreen: (At least for making DVDs. )
No KDVDs yet, until the matrix/CODEC issue is solved. The following screenshots were taken from a sample made with the procedure described above. NO FILTERS WERE USED :!: The full sample (~10MB) is here: http://www.kvcd.net/ffvfw-high-action-sample.m2v Can you find the Macroblocks :mrgreen: http://www.digitalfaq.com/archives/i.../2004/02/1.png http://www.digitalfaq.com/archives/i.../2004/02/2.png http://www.digitalfaq.com/archives/i.../2004/02/3.png -kwag |
Hi Karl!
The "constante Bitrate" thought I also had yesterday night when I switched -also when capturing using my DC30- to ffvfw "XVID" Codec .... which is fast as hell! I did played with this codec also as here also do exist almost the same options in the quantizer sections as we do use when using the mpeg2 codec. I did choose CBR 6000kbit for capturing and I did recognise that there is still a VBR quantizer based AVG output. At high quality commercials for example it grows up to 5890kbit and on a bit unsharper and letterboxed movie broadcastings it decreases by itself the AVG to almost 4000kbit :wink: My target with this codec (ffvfw mpeg2) is, to fit 3 movies in 704x576(480) on one DVDr in non-anamorph or two movies in anamorph state. So do your last try as seen above BUT now trying to encode in anamorph! state with 8px borders on each of the sides and 16px on top and bottom as overscan as you'll end up with 2 movies (each ca. 130min) on one DVD-R. I see you use "Trellis", did you recognise more quality by this?? On the old Build output got a bit worse. Inc. PS: Also a very good resolution if you will just view those KDVDs on Tv is to switch to 352x576(480) 1/2 DVD, ... I think with this encoder the results will also be awesome as the ffvfw mpeg2 ends up more sharp! Bit in such a case do also try to use std. matrix and NO DCTfilters as the stream will be horizontally streched on tv afterwards. AND nice to here that the VBV issue seems to be solved .... Ill give CBR at mpeg2 mode this night a try too! :D |
Quote:
But there are also other things included : you already see noise, but you have also sharpening (asharp, msharp, warpsharp), levels, colorspace convertion, postprocessing (deblocking / deringing)... In fact this codec can replace almost everything on your disc :-) |
FFdshow also has mplayer filters, lots of stuff to play with. The only thing it is sorely lacking is ivtc, but you could always do that from within vdub, not quite as good as decomb though. However, ffvfw compresses so well that the quality difference between 24fps and 30fps is barely noticeable at least at about 1500 kbps and up.
|
Geez, this codec rocks my socks :!: :!: 8O :D
I'm currently encoding High Crimes anamorphic for KDVD output, but still with Quality Mode. Picture Quality of the sample was awesome for the size! I'm using a Blockbuster Variance of 2.5, since it really has some nasty DCT blocks on flat surfaces even in the source. I hope I'm pleased with the results. As for the CBR mode: Will try that later tonight. :D |
Hi folks:
Hmmm... problem here :( . I guess I should test first the CBR mode to see if this corrects the problem. I'll explain myself. Encoded a 45 min sample (StarWars II) with MA script, One pass-constant quality 100, 720x576, 16:9. It took 2 hours and... but it worth the quality. No blocks on TV, and I even zoomed the image on TV and begun to see "stairs" only at 16x... 8O . BUT: during playback freezes the image from time to time :cry: . It remember me when I first tried to make SKVCD, that image froze a second and then kept playing. It looks like a high bitrate issue (but this happens at 2000-4000 kbps) or, more probably, a VBV issue. And as I read here that VBV goes down to 7... do you think that this can be the problem?. Has anybody had this problem during playback in their standalone?. I've never had this problem with KDVD before. I'll try your new advise and new settings with CBR and see if the problem solves (since it manages VBV well). BTW: what do you think is better, keep on scripting, or making use of ffvfw functions?. Don't just refer to blockbuster vs noise, but the rest: DCT, temporalsoftening, and so. |
Quote:
Quote:
Quote:
But in words of quality, ffvfw includes the same sources than avisynth plugins, so I guess the result is the same. |
Thanx Phil
:D |
Quote:
The rest of the encoding jobs and therefore the time was dedicated to ffvfw ...... .....trying, trying, trying, failure, trying, tryin, trying, SUCCESS, trying, trying, trying, failure, trying, tryin, trying, SUCCESS,trying, trying, trying, failure, trying, tryin, trying, SUCCESS,trying, trying, trying, failure, trying, tryin, trying, SUCCESS,trying, trying, trying, failure, trying, tryin, trying, SUCCESS,trying, trying, trying, failure, trying, tryin, trying, SUCCESS,trying, trying, trying, failure, trying, tryin, trying, SUCCESS,trying, trying, trying, failure, trying, tryin, trying, SUCCESS,trying, trying, trying, failure, trying, tryin, trying, SUCCESS. Means figuring out the best possible by tweaking and switching parameters. Result: :arrow: Headaik!! And I gonna cure that by watching that one effectively burned Movie tonight. :lol: PS: Yep,... my girlfriend still loves me and didn't leave me for that time taking chaos above 8) |
Quote:
Quote:
Quote:
-kwag |
Quote:
-kwag |
Quote:
-kwag |
Quote:
|
Quote:
Apparently Blockbuster is very effective with ffvfw, and as we already know, that's not the case with TMPEG in CQ mode. But still, it would be nice it ffvfw is further optimized for better quality on low dim areas, because you can clearly see the dancing blocks laughing at you most of the time :mrgreen: -kwag |
Yep, Indeed you have to rise sometimes the "variance" parameter cause that fine spray means high frequencies which will be "cut" by the Notch which does exactly perform more quantization on high frequencies and if Notch and ffvfw "get in touch" that means a lot of quantization mostly on plain parts ... by this the edges wont get that "gibbs" effect 8)
|
Kwag, you told that you remove the K-notch matrix for a while. Can you tell me what would be the expected size of the whole movie you encoded in such conditions ?
(I can't open the sample at my office and check its lenght :-)) An other way to ask the same thing : is the encoder still abble to put 3 movie on a DVD without the matrix ? |
You mean anamorphic or non-anamorphic?
|
Quote:
My suggestion for example to Jell was to drop the notch in case if the final filesize isn't big enough ;-) But he tried to fit 2 Movies on one DVD-R. But thats also the nice thing that mpeg2 ffvfw supports diff. Qant.Matrixes. Or what's your purpose Phil? |
:?: Which settings (if any) need to be changed in ffvfw to produce anamorphic? (Two movies per DVD are enough for me).
Thanks, guys, NickSteel |
You need to set AR to 16:9 in the "Output" tab. Avisynth will do the rest for you. Just make sure you check anamorphic box in MovieStacker/fitCD.
|
Thanks, Jellygoose. That's what I thought.
|
Quote:
... I don't exactly know what does this mean ( :oops: ), but he said he'll revise this issue later. Cheers |
Quote:
"If you notice, I've turned OFF "Custom quantization tables", because KVCD's notch matrix just cuts off too many frequencies with this encoder. I'll revise that some other day, when I get to talk with Milan. " I guess that he wanted to say that the quality with the matrix was too bad. But he didn't tell anything about the size of the final encoding without the matrix. My purpose isn't to put 3 movies on the disc (acutally I do 2-movie KDVD ;-)). I just want to know if the encoder still has the same efficiency without the matrix that it had with it ! Because an encoder that do better that TMPGENC but can fit only one movie on a DVD, we can find it easly ! Else, the only advantage of ffvfw will be it is free (and CCE is not, for instance). |
Hmmm even with mode set to CBR I can't seem to adjust the filesize the way kwag described it. Altering Max. Quantizer difference doesn't change the filesize at all for me. :x
Any other opinions? :roll: |
Quote:
However, if you don't have an HDTV, ffvfw with the notch matrix looks just great. Even on my HDTV, I can see some slight DCT blocks when I used the notch matrix with ffvfw, but because I had encoded with noise, the picture is really very pleasing. So on regular TVs, even large ones, this won't be seen, and you should see an excelent picture. Quote:
I don't think so. The milage should be about 4 hours without the KVCD notch matrix. With the notch matrix, expect ~7 hours ;) -kwag |
Quote:
It's very acceptable, but not as good as CCE :) Quote:
About half the space as a CCE encode, but not half as bad in quality :mrgreen: It's actually very good quality, but for some purists, it may not be :D Quote:
Quote:
-kwag |
Quote:
Quote:
In other words : "Cuts off to many..." means what on my screen, visually speaking ? (I guess is like when you push too far the DCT filter in avisynth, but I want to be sure). |
I still cannot change the filesize of my samples at all, except by lowering the Bitrate in CBR. How do you do it kwag? :roll:
|
Quote:
You'll have to do some short clip tests by yourself, with and without the matrix, to compare quality and file sizes. Quote:
But the best trade off right now, will be using the notch matrix with some noise (Blockbuster?) added, so your file size will increase but still be well below the size as if you had used the internal matrix without noise. I believe that's the perfect blend, for the moment, for using this CODEC. Maybe in the future, when the CODEC is further optimized, we can start to drop blockbuster noise, and do less filtering :) -kwag |
Quote:
As a matter of fact, I'm looking at that just right now, and it seems that the CBR value is actually sort of an "Average Bitrate" control :!: Because after you encode some footage, at least on my case, the resulting average bitrate is close to the value I set on the CBR slider, even though the MAX bitrates still peak above 3,000Kbps (possibly because I have the MIN Q set to 2 ) So it's a combination of MIN Q and CBR value, to get what you want :!: -kwag |
[quote="kwag"]
Quote:
|
Site design, images and content © 2002-2024 The Digital FAQ, www.digitalFAQ.com
Forum Software by vBulletin · Copyright © 2024 Jelsoft Enterprises Ltd.