View Full Version : ffvfw
digitize
16th February 2003, 22:50
I downloaded the newest build of ffvfw and tried to do a quick encode. I load an avs file:
#########
loadplugin("f:\avs2.5\plugins\vsfilter.dll")
opendmlsource("video-file.avi")
textsub("ssafile.ssa")
########
When i just try to do the first pass, vdub crashes, when i try vdubmod this error appears just when i click compression:
The procedure entry point WMCreateWriter could not be located in the dynamic link library WMVCore.DLL.
I can continue on to configure the codec, but when i go on to do the first pass vdubmod crashes. Im using a amd athlon tbird cpu 1.3ghz, id appreciate if anyone could help, thank you.
Edit: I've tried with previous builds, but vdub still crashes, and i've redownloaded wmvcore.dll to insure the one i had wasn't corrupted, and still vdub crashes.
digitize
17th February 2003, 06:05
Ok, I've made some progress... well not really, I no longer get the wmvcore.dll error, now it just freezes up when i try to do the first pass. Im just doing a simple divx3 (with a fourcc of mpv3) first pass encode, and vdub crashes.
digitize
17th February 2003, 17:25
Nevermind... I got help from a friend and figured out the problem :)
echo
19th February 2003, 14:24
Just a few bugs that I found in the 20030122 version:
1. During 1-pass encoding, I can't change the quantizer of B-frames whatever value I give to the b-frame offset and multiplier. They are always reported as quant 2 using ffdshow's OSD. (Quant type H.263)
2. XviD encoding produces videos that are half greyscale and half green.
3. When selecting XviD encoding, two passes 1st pass, if you click on the slider that changes the quantizer (I know I shouldn't do that and that it makes no sence anyway) VirtualDub(mod) crashes.
Anyway I like this baby with all its options. And I love that it is actually faster than XviD by about 20%. Thanks!
suni
23rd February 2003, 19:27
i still dont get the 2pass mode to work for me :o
im getting the -100 error or crashes in virtualdubmod.exe
so are there any new findings or did i overlook something obvious everybody else found on his own ?
thanks :)
Kurosu
27th February 2003, 18:38
Another bug-report, although quite small:
the quantization matrix used is always H.263, even if mpeg is selected. The bitstreams are perfectly identical, and I assume by comparing the filesize ot a xvid encoding, using H.263 quant. matrix, that it is H.263 that is always used in FFVfW.
@Suni
No idea, except for the file locking problem you surely have checked.
Tommy Carrot
1st March 2003, 01:08
I've a small question: is ffvfw really being an mpeg4 compatible codec? Because nor xvid, neither divx can decode it. And the codec's motion handling is much closer to divx311. Is it impossible that ffvfw is somewhat microsoft mpeg4 compatible?
Kurosu
1st March 2003, 01:31
I can replay ffvfw mpeg4 content (ie encoded as mpeg4) by ffdshow, xvid and divx5 (depending on the Fourcc I choose). I haven't tested the other divx3/... compliant encoding, but users have reported in that thread that it can be decoded by plain divx3.11. It's either a fourcc problem or the xvid codec installed you use.
The codec 'motion handling' is completely alike xvid (ezps, unless it has changed in the latest xvid builds), and has other motion estimation algorithms (maybe closer to divx3.11 encoder - decoder seems to support more precise motion vectors than his encoder provides). ffvfw is indeed mpeg4 compatible, as you can choose in the encoder dropdown menu divx3/ms mpeg4 v3, ms mpeg4 v1 and v2. All except 'XVID encoder' seems to use specific settings of ffvfw (motion estimation, bframes, qpel, etc).
Tommy Carrot
1st March 2003, 05:42
Sorry, but when i tried to decode the content in the virtualdub with the xvid codec (no matters which build), all i got was an error message. The same with divx503. And vice versa.
BTW, i ain't referred to the ME algorithm. Just the feeling. The blocks are much more stationary than in the xvid/divx encodings, in fact, nearly the same behaviour than divx311. I know this doesn't mean anything, but still... And i remember someone mentioned that ffvfw is divx311 compatible, not mpeg4. Maybe just my memory playing with me. :)
digitize
1st March 2003, 06:48
@Tommy Carrot
I think that you read that it can create divx3 compatible streams, not that it only creates divx3 compatible streams.
ookzDVD
11th March 2003, 10:20
2pass- 2nd pass trick,
Please close the Vdub and retry again the 2nd-pass should working :)
easyfab
15th March 2003, 18:05
There is some changes in cvs.
Can someone make a build (athos :) )?
Thanks
athos
16th March 2003, 22:20
Originally posted by easyfab
There is some changes in cvs.
Can someone make a build (athos :) )?
Thanks
Your wish is my command. A new build has been mailed to nitrogen.
ChangeLog:
2003-03-15 22:39 milan_cutka
wmv9 decoding (at that's all for today)
2003-03-15 22:02 milan_cutka
fixed extradata, WMV1,WMV2 decoded by libavcodec, working on wmv9
decoding
2003-03-15 17:05 milan_cutka
wmv codec interface support for encoding (wmv9 in AVI)
2003-03-14 08:40 milan_cutka
updated vc6 project
2003-03-14 08:10 milan_cutka
no message
2003-03-14 06:36 milan_cutka
many bugfixes, few new features, ...
2003-03-13 15:57 milan_cutka
no message
ookzDVD
17th March 2003, 09:44
@athos,
I think nitrogen is still "off-air" :)
do you mind if you upload the build to your website ?
Thank you.
kilg0r3
17th March 2003, 10:01
Has anybody made any serious quality comparisons between xvid and ffvfm already?
athos
17th March 2003, 11:38
Originally posted by ookzDVD
@athos,
I think nitrogen is still "off-air" :)
do you mind if you upload the build to your website ?
Thank you.
Done.
ookzDVD
17th March 2003, 12:02
@athos,
Thank you.
PS.
which one ?
I can't find it on the http://athos.web1000.com/ffvfw.html
athos
17th March 2003, 13:10
Sorry, forgot to update the html. Should work now, the direct link (i dont think it works since the webserver checks for referrer) is http://athos.web1000.com/ffvfw-20030316.exe
Uli
17th March 2003, 13:59
Originally posted by ookzDVD
@athos,
...
http://athos.web1000.com/ffvfw.html
I can't get it :(
I always get redirected to
http://download.startsurfing.com/index.html?affid=global
which wants me to download some dam'n privacy stuff?
greetz, Uli
sam_b
17th March 2003, 14:07
Right, this is getting silly, so (assuming it's OK with athos) here's a mirror:
http://dubya.plus.com/ffvfw-20030316.exe
Don't y'all go burning my ADSL connection :D
kilg0r3
17th March 2003, 14:11
It's too large it seems.
We are sorry, but the file you are trying to download is larger than the maximum length permitted under current network conditions. Please try again later when network conditions change. The file you were trying to download was 531884 bytes.
Uli
17th March 2003, 18:40
Originally posted by sam_b
...
Don't y'all go burning my ADSL connection :D
gee, thanks ;)
greetz, Uli
easyfab
17th March 2003, 19:06
Thanks athos,
I've made some little tests:
- the 2 pass works correctly again :) (on a 2min test file)
- wm9 directly in avi don't work for me (some more tests to do)
Kurosu
17th March 2003, 21:43
On a Duron 700 (not a Morgan core, therefore), the codec provokes the crash of any application that loads its dll. It seems the CPU features aren't inspected well, or that an invalid instruction is used (like SSE code). Build from 9th of February works ok.
On a side not, what are the requirements for building by oneself the codec? I suspect:
- cygwin with gcc, binutils, make and mingw runtime (from ffdshow)
- Visual C++ for final build
Buy then, what are the steps for building it?
athos
17th March 2003, 21:59
Essentially, you gcc "make" the parts that are gcc, and then build the VC parts.
But you need some more stuff, like Nasm, DirectX SDK (for DirectShow filters), Windows Media SDK and Samples (for WM9), STLPort, and of course NSIS for making the installer. There might be more that I cant think of right now.
ookzDVD
18th March 2003, 11:21
Originally posted by easyfab
[B]Thanks athos,
I've made some little tests:
- the 2 pass works correctly again :) (on a 2min test file)
I just test the 2 pass,
the 1st-pass was OK, but the 2nd-pass was error.
I have to close the VirtualDubMod and restart the 2nd-pass again
to make it works :(
Koke
18th March 2003, 13:37
Here it`s OK. Both jobs went fine (whole movie)
in VDubMod.
NiTroGen
18th March 2003, 19:07
OK, the new build is on my site, too. Sorry for the delay. I had a trip this weekend. How should I know that Athos made a new build? :(
athos
18th March 2003, 19:49
Originally posted by NiTroGen
OK, the new build is on my site, too. Sorry for the delay. I had a trip this weekend. How should I know that Athos made a new build? :(
Hey, I almost didnt know it myself ;) I check the cvs regularly, but I missed this long-awaited update. Tried ffdshow too but it wont compile. Well, I hope milan will check in here soon so he can see how much his work is appreciated.
Calculon
18th March 2003, 20:54
"- the 2 pass works correctly again"
Same for me. I set VDubMod to encode Star Trek IV which is about 2 hours long while I was at school and it worked like a charm...
kilg0r3
18th March 2003, 21:29
I don't want to nag anybody here but what about quality comparisons?
unplugged
19th March 2003, 01:31
Excuse me, I'm interested for first time to ffvfw,
where can we get recent binaries?
Nitrogen, your web pages seem down since severals days... :(
NiTroGen
19th March 2003, 03:52
Originally posted by unplugged
Excuse me, I'm interested for first time to ffvfw,
where can we get recent binaries?
Nitrogen, your web pages seem down since severals days... :( I've just checked them and they seem fine. Did you try Athos' web page?
EDIT: Check the new ffvfw site: http://nitrogen.kicks-ass.net/ffvfw It's on a new server. It should be faster, but it may be offline sometimes.
ookzDVD
19th March 2003, 03:58
@nitrogen,
Your website is OK, I can open it and it's updated.
Thank you.
@forum,
2 pass with b-frame is still broken.
2 pass without b-frame is OK!
Koke
19th March 2003, 06:04
2 pass with B frames (3) is OK here.
Avisynth 2.5 + VDubMod with script:
LoadPlugin("E:\Program Files\AviSynth2\PluginYV12\unfilter.dll")
LoadPlugin("E:\Program Files\AviSynth2\pluginYV12\decomb.dll")
LoadPlugin("E:\Program Files\AviSynth2\pluginYV12\mpeg2dec3.dll")
mpeg2source("D:\VOBs\john q\johnq.d2v")
crop(16,16,-16,-16)
lumafilter(-1)
unfilter(-5,-5)
BicubicResize(608,336)
levels(0,1.165,241,46,246)
unfilter(70,70)
Kamui-Dash
23rd March 2003, 02:30
Hmm, I dont see Encoder: Divx3 on the latest ffvfw-20030316.exe build anymore. Can sum1 help?
raistlin2k
24th March 2003, 13:04
using m2a px3 0.16c:
Can't encode a video with 720x312 pixels, found the reason 312 is not a multiple of 16, changing to 320 works perfectly.
WHY does ffvfw require a resolution dividable through 16, DivX & XVId do not??
Could you fix this, please?
Thanks
Raist
sam_b
24th March 2003, 13:12
You should always use multiples of 16, even with divx and xvid, for various reasons.
trbarry
24th March 2003, 14:24
But you need some more stuff, like Nasm, DirectX SDK (for DirectShow filters), Windows Media SDK and Samples (for WM9), STLPort, and of course NSIS for making the installer. There might be more that I cant think of right now.
Is there WM9 support in ffvfw now?
- Tom
Kamui-Dash
24th March 2003, 14:59
The newest Lib codec doesnt show up in vdub, could sum1 help me?
bill_baroud
24th March 2003, 16:35
@TrBarry:
latest changelog says
2003-03-15 22:39 milan_cutka
wmv9 decoding (at that's all for today)
2003-03-15 17:05 milan_cutka
wmv codec interface support for encoding (wmv9 in AVI)
but i suppose you already know it
raistlin2k
24th March 2003, 17:19
You should always use multiples of 16, even with divx and xvid, for various reasons
Which reasons?
I encoded more than 180 movies, quite often with 720x312, no problems at all, so why should I use multiples of 16. I don't write this to offend anyone, I just really want to know since I always want to improve my encodings!
Thanks
Raist
P.S. Although I should perhaps always use multiples of 16, could it be arranged that I can select ffvfw from Compressor-choose-windows even if I'm encoding with other resolutions? Thanks
Kurosu
24th March 2003, 18:30
why should I use multiples of 16.
Because MPEG encoders divide the picture in blocks of 8x8 (rather, 16x16 for luma), and performs most operations on such units. Using non MOD16 resolutions needs a more difficult handling (that's why sometimes the output exhibits strange artifacts) and some kind of 'padding'. See it as inefficient as if leaving black borders not on MOD16 boundaries. Only supporting MOD16 resolution is then maybe a choice to simplify the coding.
If it's a matter of encoded aspect ratio, 720/312 ~ =2.307
The official SAR is normally 2.35, so 720/304 ~ = 2.368 is much closer
(should be 306 to be closer to 2.35). But are you really able to make the difference between such close aspect ratios?
could it be arranged that I can select ffvfw from Compressor-choose-windows even if I'm encoding with other resolutions?
I don't think so. Using MOD16 resolution enables to make assumption on data alignement and data size. Forcing non MOD16 resolution into ffvfw through some kind of hack would probably lead to memory leaks and alignement exceptions.
sam_b
24th March 2003, 19:43
And graphics card drivers can be buggy (on playback) with non-mod16 resolutions. Some don't even like non-mod32 resolutions horizontally.
athos
24th March 2003, 20:24
Originally posted by trbarry
Is there WM9 support in ffvfw now?
- Tom
There is, but I'm not sure if or how well it works. I know milan is working on it though.
sungey
24th March 2003, 23:45
if you are encoding using Xvid for Divx playback ... u need to use mod 32 for width and mod 16 for height ... height non mod 16 (e.g mod 8) will produce artifacts ...
sungey
24th March 2003, 23:50
one question bout ffvfw Divx3 encode ... in Nandub .. there's an option called crispiness modulation which change the crispiness of the frame according to the motion (it masks the blockiness in high motion better)... does ffvfw does the same thing ?... if my eyes didnt fail me .. i think high motion in ffvfw divx3 isnt masked as good as nandub's encode ... any idea ?
athos
26th March 2003, 00:38
New build up @ http://athos.web1000.com/ffvfw.html
Also emailed it to nitrogen.
Changelog:
2003-03-25 08:55 milan_cutka
updated vc6 project
2003-03-25 06:55 milan_cutka
more mpeg2enc controls, strip removed from nsi, assume interlaced
if dy>... option
2003-03-24 09:34 milan_cutka
some vflip issues
2003-03-24 06:56 milan_cutka
support for more colorspaces when decoding, changed the way how the
graph works (again), few fixes
2003-03-21 14:57 milan_cutka
no message
2003-03-21 08:58 milan_cutka
vc6 project update, crashes on exit :(
2003-03-21 06:38 milan_cutka
aspect ratio selection (needs more work), AVISynth script playing
using fake avi file (just a proof of concept now)
2003-03-20 15:26 milan_cutka
cosmetic fixes, dshow filter can connect directly to file writer
filer (or similar)
2003-03-20 08:37 milan_cutka
vc6 project update
2003-03-20 06:44 milan_cutka
mpeg2enc two pass encoding possible, working on mpeg2enc config
pages
2003-03-19 15:03 milan_cutka
trying to design mpeg2enc configuration pages
2003-03-19 08:27 milan_cutka
vc6 projects update
2003-03-19 06:47 milan_cutka
an _attempt_ to use mpeg2enc in ffvfw (WIP)
2003-03-19 06:45 milan_cutka
libavcodec modifications: don't obey qmin/qmax in RcOverride
sections, compute frame average quantizer
2003-03-16 21:58 milan_cutka
few fixes, libavcodec hack: adaptive quantization in fixed quant
mode
This time I have enabled ff_theora (was in another project) and milans new mpeg2enc. Don't know if either work.
NiTroGen
26th March 2003, 04:14
Thanks, Athos. I've uploaded it to my site (http://homepages.pathfinder.gr/ffvfw) and to its faster (but not always online) mirror (http://nitrogen.kicks-ass.net/ffvfw).
slavickas
26th March 2003, 16:28
hi
how to playback theora ?
and why it needs debug version of msvcrt?
vBulletin® v3.8.11, Copyright ©2000-2026, vBulletin Solutions Inc.