ffvfw VIDEO CODEC
Continued from ffmpeg thread...
|
So lets continue ;)
Kwag, you mentioned that earlier test with ffvfw ended with crash at all time. I was preforming (or should i say try to preform) two pass encode with VDMod. Result -> crash. Did you try this? |
Geez,
Then although it's using libavcodec, tmpgenc is faster than ffvfw :?: :?: :?: Damn :!: Ok then I don't see any point on going much further with our efforts :cry: unless Karl sees any brand new potential that could blow tmpgenc or cce out of the skies. Anyway it's been fun to learn how to use this one. Karl, u da man. What do we do :?: Should we keep running more tests? Should we wait for Milan answer? Regards |
Quote:
In one pass, it does work fine, but with sky high bitrates. Even at 352x240, I got bitrates as high as around 3,000Kbps, which is too much. -kwag |
Hi rds_correia,
I really like the CODEC, and I like the motion estimation even better that TMPEG's. I wonder how high the bitrates will go doing an MPEG-2 at 704x480 :idea: I'm about to try that, because maybe it can be used to make KDVDS. It would be good to be able to patch the matrix though :roll: Let's see if the developer answers the E-Mails, or even joins us here, and we can start a new storm :mrgreen: -kwag |
ffvfw is faster than TMPG :!:
Maybe I did'n clearly explained this. Sample movie was encoded in 41sek(ffvfw) and 64sek(TMPG). 41<64 ;) Seek acurate to highest level increes time for ffvfw to ~50sek, buth i dont know all setings. Lets wait for Milans reply. Thats my opinion. As always Kwag is faster than bullet ;) |
Ok let's give it a try because maybe we haven't seen enough from this baby.
Anyway about the matrix, I already patched the matrix successfully as I said in the mencoder thread. I even mentioned the steps... Can u try that too? Cheers |
Quote:
Currently doing a test encode :cool: -kwag |
Yes. I can change matrixes the way you described.
It's 0:40 i my time zone, and i must go up realy early :x Have a nice talk :) |
Well it's 23:42 in my zone and I must get up at 06:20 but...
You know, just gotta make this last statement: I can't patch the correct intra matrix on it... the last two lines of it should be: 29 34 37 38 40 48 58 69 34 37 38 40 48 58 69 79 and it's changing it to 29 34 37 38 40 48 58 75 34 37 38 40 48 58 75 83 sniff, what will we do?? Milan boy, where are you???? |
Hey kwag,
post your results :twisted: |
Get ready for the results .......... ...... ........ :mrgreen: :mrgreen:
|
Man!
You should be the presenter of the Movie Academy Awards :twisted: lolol |
Hit me with quality, speed, etc..
Maybe a comparison with tmpgenc?? |
LOL :D
Here it is: http://www.kvcd.net/ffvfw-red-planet.test.m2v CODEC ffvfw with KVCD "Notch" matrix, using constant quality set to 100 :!: Max bitrate goes up to 1,779Kbps and average bitrate of clip is only 892Kbps 8O :mrgreen: NO FILTERS USED :!: Plain .avs only to resize to 704x480 with 2 blocks-TV overscan, so I guess with the MA script, this will be very promising :lol: Edit: Sample is MPEG-2 with a max GOP size of 18 frames. -kwag |
Fedback. Feeback. Feedback. Feedback.
:mrgreen: :mrgreen: :mrgreen: :mrgreen: |
8O
I've got nothing to compare about quality but it sure looks great. What about speed? Have you noticed any problems with your patched matrix as I did?? |
Great quality, now that i've finally finished downloading it!!!
|
Man,
I gotta crash. Otherwise I might put my job in danger :lol: See you guys tomorrow. Hoppe I can run some tests also. Dunno how to patch that matrix perfectly yet though... C ya |
Later ;)
|
Quote:
I guess the KVCD matrix cuts off a lot of frequencies ( with this CODEC ). So I increased the value to 100, and it looks great. To increase the MAX (and MIN ) bitrates, you can lower the MAX Q value (defaults to 31) so your average bitrate will increase. So we can do a "sort of" prediction a-la-CCE by lowering the MAX quantization value, and leaving the quality set to 100%. -kwag |
Hi guys,
I'm affraid I didn't understand a word on Kwag's last post :oops: I'm not a super noob anymore, but I'm definitely still a noob :oops: Anyway, any news from Milan so far? He sure could give us some help with this. Let's see if I get home early to run some more testing. Cheers |
Don't be afraid Correia, Kwag is not speek in english :D joke ;)
I must say that quality is great, but this sample is pretty dark, so I think that is the reason of low bitrate. I have no comparition with other MPEG2 samples while i only produce MPEG1 clips. So if we have that small MAX bitrate with 100CQ then maybe two pass encoding (when this start to work) may be very helpfull. We will make the most of full 2500 (or 2300) bitrate -> ergo, even beter quality, and more predictive file size. |
I also thought that the image was quite dark but anyway there are pretty nasty parts in the source (comet fire) that the encoder treated like a charm even without any kind of filtering.
Could u guys please post your procedure and tools involved in your trials? Thanks |
Hello my friends :D
Last night I encoded the complete "Red Planet" movie, KDVD (704x480 + 2 overscan blocks), with the MA script. To my surprise, the total video size was ~800MB (less audio) 8O The maximum bitrate peaks for this movie is 6,664Kbps, and the average bitrate is 1,216Kbps 8O The reported Q factor given by Bitrate viewer is 3.34, and is constant throughout the complete movie. There was just no way to get my wanted sampler size of ~20MB, and the largest size was ~17MB. Even after setting MIN and MAX Q values to 2 (maximum quality). The matrix is really making this CODEC compress the crap out of the movie :lol: But the resulting quality is just amazing :!: Here are the VirtualDub screenshots (in sequence) for your reference, on the CODEC parameters I used, followed by actual screenshots of the full encode: http://www.digitalfaq.com/archives/i.../2004/01/1.jpg http://www.digitalfaq.com/archives/i.../2004/01/2.jpg http://www.digitalfaq.com/archives/i.../2004/01/3.jpg http://www.digitalfaq.com/archives/i.../2004/01/4.jpg http://www.digitalfaq.com/archives/i.../2004/01/5.jpg http://www.digitalfaq.com/archives/i.../2004/01/6.jpg http://www.digitalfaq.com/archives/i.../2004/01/7.jpg http://www.digitalfaq.com/archives/i.../2004/01/8.jpg http://www.digitalfaq.com/archives/i.../2004/01/9.jpg http://www.digitalfaq.com/archives/i.../2004/01/2.png http://www.digitalfaq.com/archives/i.../2004/01/3.png http://www.digitalfaq.com/archives/i.../2004/01/4.png With this, for me, it's the end of file size prediction :mrgreen: Note: I had to change the last three values of the matrix, because the codec works on 8 bit values and can't go over a value of 63. You'll notice that in the right bottom side of the matrix. Anyway, the quality is just AWESOME, and this CODEC blew away TMPEG (and every other encoder I have tried :!: ) in motion estimation. Look at the screenshots, and you'll see that there are BARELY visible atrifacts(mosquitoes) around objects, specially in the last shot, around the small objects. Note 2: The produced .mpg file must be demuxed, and then run PULLDOWN.EXE on the file. This is for progressive sources. Also, you must use DVDPatcher to patch the first header of the file to some bitrate, like 8,000Kbps. The encoded MPEG file reports a bitrate over the DVD standard, and DVD authoring programs won't import the file. That's all you have to do. I imported my file into TMPGEnc DVD author, after this process, and worked like a charm :) Enjoy :!: -kwag |
GREAT :angel: :hammer:
:ole: Results are amazing :) I don't fully understand what you mean by " it's the end of file size prediction". We have not force two pas engine to work, and with one Qpass, we still don't know target file size, are we? Edit: I found this in short help files attached to instalation package: Quote:
This files you can find in \ffvfw\help. |
Quote:
This means that I can put an average of about 7 hours of footage (movies) on a single DVD with this quality (just like the sample I posted) This is almost full D-1 (704x is almost the same as 720x in quality) So this will now allow 3 or 4 full run length movies, without having to do any file size prediction, and that is sweet :cool: -kwag |
8O 8O 8O 8O 8O
Quote:
I just flew over this thread "now", cause a few hours ago I thought someone gots again a codec problem :lol: But NOW: Inc. = 8O Is that a new release of ffvfw including that mpeg2 option or something extra to install. BECAUSE Tonight I will be at home, girlfriend is in the house of her parents, AND SO I want to play with something! :lol: |
@inc:
Think we missed alot. Have a look here: http://www.kvcd.net/forum/viewtopic....r=asc&start=35 Just saw those jumping smileys... @Kwag I don't have the Mpeg2 Encoding option, do you know what's missing? I can select mpg2enc MPEG2... |
I'm "preparing da engines"!! :D 8)
|
Same than Inc and Krassi.
But I already have a question : in the screenshot we can see "encoder MPEG2" and "output to file : mpeg1 stream". What is the format in the end ? MPEG1 or MPEG2 ??? |
Karrasi - you should use only MPEG2 (without mpeg2enc). If its missing, try reinstal ffvfw.
Dialhot - at the end we have MPEG2 video stream. ffvfw with MPEG1 stream option produces MPEG1 or MPEG2 streams (depend on choosen encoder). |
@Hydeus
Thanks. I would like to select it, but i don't have this option. Maybe a missing codec :?: |
All codecs are in one package. Try reinstal.
I must go now. Kwag will probably explain everything beter than me :lol: And we still waitnig for answer from Milan (developer of ffvfw). There are few bugs. |
I just checked KWAGs ffvfw sample .... seems that it takes ca. 900kbit as average, really nice at 704x480!
Christmas I did test on TmpgEnc a new GOP and Matrix on that sequence where a avg Bitrate of 500kbit @ 480x576 came out. http://home.arcor.de/the.incredible/sample.mpg (do preview using Vdub!) Also nice, so this evening I will see how do my settings will harm with that ffvfw engine! One hour and I'll be at home!! :arrow: :D But where I can set the GOP at these ffvfw settings?? Well I'll figure it out. EDIT: http://www.digitalfaq.com/archives/i.../2004/01/4.jpg Aha! "Maximum I frame interval" means max frames in a gop ?? "Minimum I frame interval" means max reducable I frame interval in case of like a scene change detection??? So if I set it also to 15 (PAL) would mean fixed GOP lenght? "max consecutive b frames" means the count of b frames between an I and a P frame ??? |
A reinstall didn't help :(
So this is what i've installed: ffdshow-20020617.exe ffvfw-20030415.exe mencoder-20031126.exe |
Quote:
You'll see it when you demux the MPEG with TMPEG. :) BTW, the quality when you select MPEG-1 instead of MPEG-2 is lower :!: So this encoder seems to do a better job on MPEG-2 than in MPEG-1, so it's not as good as TMPEG for doing low bitrates. But for KDVD, it rocks :mrgreen: -kwag |
Quote:
-kwag |
Quote:
Quote:
Quote:
Quote:
-kwag |
I've found my error :D
I've downloaded the latest ffvfw from sh0dan's link: http://cultact-server.novi.dk/kpo/avisynth/avs_cvs.html Now i can select it 8) |
Site design, images and content © 2002-2024 The Digital FAQ, www.digitalFAQ.com
Forum Software by vBulletin · Copyright © 2024 Jelsoft Enterprises Ltd.