ffvfw

[sh0dan]

@milan: Sorry about the older libraries - I just copied the libraries from the last build, because I didn't know where to get them. I don't have access to the system that compiles this, as I'm on my backup computer at home.
I think someone with the proper setup should do a binary - I'll update the one on the site, when you post it.

[Kamui-Dash]

This is pretty dumb but I was wondering if any1 got 2pass working from jobcontrol? I usually gets PARSE ERROR. Does any1 knows what this means?

[grabengulli]

I try to compile the libavcodec_enc.dll but i have had an error:

ffvfw\src\ffmpeg\libavcodec\ratecontrol.c(131) : error C2520: conversion from unsigned __int64 to double not implemented, use signed __int64

what can i do?

thanks,
GrGu

[athos]

Originally posted by grabengulli
I try to compile the libavcodec_enc.dll but i have had an error:

ffvfw\src\ffmpeg\libavcodec\ratecontrol.c(131) : error C2520: conversion from unsigned __int64 to double not implemented, use signed __int64

what can i do?

thanks,
GrGu
Do you compile with gcc? what version?
I suggest you use gcc for this (libavcodec, mplayer and libmpg2), as parts of the code is hand-optimized in assembler. If you use VC++, these parts will be unoptimized, or rather only the usual automatic optimizations by the compiler. I have compared using gcc and intel c++ compiler in vc++, and gcc is a lot faster thanks to the assembly parts.

[grabengulli]

I have tried it with gcc and all works fine. Thanks.

Grgu

[milan]

@kempfand
I wanted that registry entry to be added by installer, but then ffvfw was reported twice by VirtualDub and other VFW software so I decided to not do this.

@sungey
Stats files produced by ffvfw should be compatible with those produced by XviD. However ffvfw doesn't fill kblk, mblk and ublk fields. If needed I could fix that.

@grabengulli
Which Visual C++ version are you using? I just tested it with VC6+SP5+Processor pack and it compiles fine. Of course GCC is preferred compiler for libavcodec, but I want it to be compileable using VC too.

@Bach, @Kamui-Dash
I'll check that immediately.

[milan]

@Bach
I'm not able to get that error. Please try to capture debug data using DebugView (www.sysinternals.com), maybe there will be something useful.

@Kamui-Dash
Which version of VirtualDub are you using? ffvfw state data are quite long (usually about 1500 bytes) and maybe not every VirtualDub version is able to handle that. I just tested VirtualDubMod 1.4.13 and it works fine.

BTW I updated ffvdub and had similar problem storing the filter data to allow scripting support. The buffer supplied by VirtualDubMod was not big enough.

[Blight]

milan:
are you keeping tabs on the ffdshow thread? there are a few bugs I've reported that could use resolving.

[milan]

I'm now focused more on ffvfw. I know about ffdshow issues which need to be solved, but I'd like to do that after ffvfw release.

[Kamui-Dash]

I used a vdub ver where i first copied the vdub files in nandub folder than vdubmod in it afterwards. The problem has been solved after I installed vdub in its own folder. :D

[JasonFly]

What does really mean Divx3 when choosing the encoder setting.
Does it mean that the movie will be encoded whith ffvfw and will only have a div3 fourcc or does it mean that you can encode in divx3 with ffvfw?

[milan]

When selecting Divx3 ffvfw uses libavcodec to encode video to DivX 3 compatible format. You should be able to decode it by the original DivX 3.11 codec. Same goes to MSMPEG4v2, however older versions of libavcodec produced buggy streams.

[grabengulli]

Originally posted by milan
@grabengulli
Which Visual C++ version are you using? I just tested it with VC6+SP5+Processor pack and it compiles fine. Of course GCC is preferred compiler for libavcodec, but I want it to be compileable using VC too.


I use VC6+SP5 but without the Processor pack . Only with MASM & NASM.

[milan]

Then try to install the processor pack or at line 69 of ffvfw/src/ffmpeg/libavcodec/common.h (applies to current CVS) change

# define uint64_t_C(c) (c ## ui64)

to

# define uint64_t_C(c) (c ## i64)

However I don't know how it will affect the encoding, but I guess it should be OK. Even in ffmpeg CVS the i64 suffix is present, I changed it only in ffvfw CVS copy of libavcodec.

[LigH]

Nitrogen's mirror seems to be dead, but the greek Pathfinder is still alive:

http://homepages.pathfinder.gr/ffvfw

[unixfs]

Hi,

using ffvfw-20030325.exe (15/4 release crashes when using mpeg2enc)
I made a couple of short two-pass encoding of a black and white movie and a cartoon, using mpeg2enc-mpg1 and the following settings:

resolution: 352x288, search-radius 32, 4x4: 2, 2x2: 2, no alt-curve, payback proportionally, 772 kb/s, output to an external file.

I compared the output of both with the corresponding encodings made using libavcodec's MPEG1; to me it seems that mpeg2enc's results are
quite superior, but are much slower (roughly 9 fps during both passes on my Athlon XP 1500+). Surely it's a good solution for making VCDs without using TMPG.

@milan: as far as I know mpeg2enc doesn't support 2-pass encoding natively, is it a feature that you added? It's great!.

Can it be made faster? :)

Thanks very much for you excellent work.

[milan]

I didn't performed any comparison of libavcodec mpeg1 encoding and mpeg2enc. I added mpeg2enc because I wanted to have an MPEG2 encoder in ffvfw and when Nic released his mpeg2enc compile I tested it and it seemed quite good to me.

I don't know if I can make it faster. Maybe it'd be useful to replace mpeg2enc quantization and DCT routines with those from libavcodec, but I don't think I'm capable of doing this now.

mpeg2enc two pass encoding uses the XviD two pass code. I find it very flexible and using it for Theora too. However this code doesn't have option for setting minimum bitrate which would be probably useful for MPEG1/2 encoding.
In MJPEGtools CVS log I saw that the mpeg2enc developers are implementing their own two pass encoding routines.

[sungey]

EDIT : wrong forum ... sorry

[grabengulli]

Originally posted by milan
Then try to install the processor pack or at line 69 of ffvfw/src/ffmpeg/libavcodec/common.h (applies to current CVS) change

# define uint64_t_C(c) (c ## ui64)

to

# define uint64_t_C(c) (c ## i64)

However I don't know how it will affect the encoding, but I guess it should be OK. Even in ffmpeg CVS the i64 suffix is present, I changed it only in ffvfw CVS copy of libavcodec.

I have tried it but it dosn't work. I have had the same error like before. I think this is an error at Visual Studio on my PC and not in the Source Code, because yesterday I had the same error in an other Source Code.

GrGu

[downloada]

hi,

i don't know if this has been posted in this thread before (if it's been, than sorry!), but i have a problem with the 2003-03-25-version of ffvfw.

since i installed that version, everytime i open the "compression"-dialog in vdub/vdubmod, an error message saying msvcrtd.dll was missing pops up twice, the same thing happens when clicking on ffvfw in that dialog. i don't have that .dll on my pc and i couldn't find it on the net, so i don't know what would happen if i put that dll in my windir.

i didn't encode with it (just installed it because it was needed for one ffdshow-version), but it's impossible to do an automated encode via gordianknot with this errormessage, because i have to click it away twice. for now, i have uninstalled it and everything works fine, so it has to be a problem with ffvfw.

cu

[Assault]

@ downloada
A simple search in google returned that ;) http://freespace.virgin.net/nick.brandon/msvcrtd/

Regards
Assault

[athos]

New build up today @ http://athos.leffe.dnsalias.com/ffvfw-20030502.exe

Changelog since my last build (2003-03-25):
2003-05-01 07:53 milan_cutka

no message

2003-04-30 14:56 milan_cutka

fixing AviSynth script in AVI

2003-04-30 13:53 milan_cutka

xvid doesn't support selecting qmin/qmax separately for I,P frames
in CBR mode - GUI updated

2003-04-30 12:24 milan_cutka

AviSynth and uncompressed video only installation package

2003-04-30 09:10 milan_cutka

outdated

2003-04-30 09:09 milan_cutka

fixed some dialog flickering when changing codecs and modes

2003-04-30 08:25 milan_cutka

no message

2003-04-30 07:03 milan_cutka

separate quantizer min/max for I,P,B frame types in libavcodec,
redesigned quantizer dialog, working on xvid two pass code to
support multipass encoding

2003-04-25 14:08 milan_cutka

no message

2003-04-25 09:29 milan_cutka

fixed memory leaks in mpeg2enc (>20MB), cosmetic

2003-04-24 15:37 milan_cutka

fixed changing of encoder without closing the application

2003-04-24 15:10 milan_cutka

no message

2003-04-24 14:46 milan_cutka

no message

2003-04-24 14:32 milan_cutka

reduce huge fpsRate and fpsScale value

2003-04-24 12:35 milan_cutka

encoding to uncompressed video, dialog creation cleanup

2003-04-24 07:19 milan_cutka

updated VS.NET project, cosmetic

2003-04-23 19:01 milan_cutka

cleanup

2003-04-23 18:35 milan_cutka

storing valid kblks,mblks,ublks data to xvid two pass code stats
file

2003-04-23 15:04 milan_cutka

OGM muxer

2003-04-23 06:58 milan_cutka

working on sound support

2003-04-22 13:11 milan_cutka

I'm able to get some sound from it, crashes of exit

2003-04-22 08:11 milan_cutka

vc6 project update

2003-04-22 06:22 milan_cutka

support for quantizer bias, working on sound support in makeAVIS,
cosmetic changes on stats page

2003-04-16 06:27 milan_cutka

cosmetic

2003-04-15 09:49 milan_cutka

also look in ffvfw.dll folder for dynamic libraries

2003-04-15 06:47 milan_cutka

mb_stride

2003-04-14 14:24 milan_cutka

headers cleanup

2003-04-14 06:53 milan_cutka

WMV9 Screen codec in AVI support, HuffYUV (YUY2) encoding,
XVID_CSP_YV12 input for XviD encoding (bypasses bug in XviD
XVID_CSP_USER input), generic support for multiple colorspaces for
encoding, can be compiled using GCC (omits AviSynth colorspace
converters), stores/loads ffdshow settings in ffvfw state data

2003-04-12 06:59 milan_cutka

fixes in colorspace conversion

2003-04-11 15:45 milan_cutka

XVID_CSP_YV12, XVID_CSP_I420 -> XVID_CSP_USER conversion cleanup,
fixed forcing of input colorspace

2003-04-11 12:08 milan_cutka

using Tconvert

2003-04-11 07:53 milan_cutka

working on more generic colorspace conversion class, psnr display
for libavcodec and xvid, raw video decoding, updated libavcodec

2003-04-11 07:50 milan_cutka

working on more generic colorspace conversion class, psnr display
for libavcodec and xvid, raw video decoding, updated libavcodec

2003-04-08 15:55 milan_cutka

working on psnr stats

2003-04-08 14:40 milan_cutka

no message

2003-04-08 14:30 milan_cutka

updated libavcodec, allow mb quant limits only when adaptive
quantization enabled

2003-04-08 12:08 milan_cutka

video properties in makeAVIS

2003-04-08 10:27 milan_cutka

fixed output to YUY2 (and generally flipped) colorspaces, forcing
interlacing in makeAVIS (break AVIS backward compatibility)

2003-04-03 09:04 milan_cutka

trying to compile makeAVIS with mingw

2003-04-03 06:44 milan_cutka

updated makefiles

2003-04-02 16:51 milan_cutka

fake AVIs can force output colorspace

2003-04-02 15:12 milan_cutka

no message

2003-04-02 15:09 milan_cutka

no message

2003-04-02 14:31 milan_cutka

YUY2 -> YV12 fix in decoder, updated libavcodec, updated GUI to
better reflect libavcodec features

2003-04-02 06:45 milan_cutka

no message

2003-04-01 15:24 milan_cutka

no message

2003-04-01 12:29 milan_cutka

fixed crash when wmv9 wasn't available

2003-04-01 11:32 milan_cutka

load wmvcore.dll at runtime

2003-04-01 10:15 milan_cutka

updated libavcodec, WMV codecs FOURCC autodetection

2003-04-01 06:46 milan_cutka

cleanups and fixes

2003-03-28 10:45 milan_cutka

updated vc6 project

2003-03-28 06:48 milan_cutka

improved Theora supports, always show disabled checkboxes and radio
as unchecked, IffProc3, changed dll search order, warning fixes

2003-03-27 16:34 milan_cutka

va_end

2003-03-27 16:33 milan_cutka

no message

2003-03-27 10:08 milan_cutka

more precise sup_trellistQuant

2003-03-27 06:38 milan_cutka

mostly just cosmetic changes

2003-03-26 14:49 milan_cutka

,

2003-03-26 14:42 milan_cutka

more detailed version info (not yet for libswscale.dll)

2003-03-26 10:53 milan_cutka

let's call it "multipass encoding" (only for libavcodec)

[LigH]

Thank you for this new build: I was just about to ask why I was unable to select several encoding models (e.g. even MPEG-4) in previous versions; but with this build, they all above "Theora" are available again (in previous versions, the list started with Theora).

[sungey]

thanks for the new build

[slxslxslx]

can not download the new build
:(

[NiTroGen]

Originally posted by slxslxslx
can not download the new build
:( Try to download it from my site(s)...

[aliens3]

I have some video clips encoden in wm8 and when i discovered that ffvfw can decode wmv1 and wmv2 I was happy..
Unfortunately, wmv1 works ok but not wmv2. when I open in vdub an avi (created with asftools) everything is weird.. green and pink..wmv1 works ok, still..

[MystSith]

In ffvfw, there is an "image preprocessing" with ffdshow image filters.
How to use it?

[moonman]

Mystsith:
If you mean to use ffdshow to do postprocessing you should do this:

You must have both ffvfw and ffdshow installed. Then in vdub mod, select ffvfw as codec and under configure choose witch encoder such as xvid for an example. then make your settings for that encoder. Under "Input" settings enable "image processing" and then go in to "configure" now you are in ffdshow. There you can make all your postprocessings you want for example resizing. Many of the filters ffdshow uses is from avisynth but built in. You can even use a avisynth script there if you want. this is the way i encode mpeg4 and i think it is the easiest way. Make sure you have the most recent versions of ffdshow and ffvfw.
moonman

[Uli]

Hello,

installing the latest build of ffvfv (ffvfw-20030502.exe) on Win2K running P4 Willamette 1.8 GHz stops VirtualDub and VeeDubP4 from working. Uninstalling ffvfv solves the problem. The processes of VirtualDub or VeeDubP4 are starting at 99% CPU load and keep hanging until killed with task manager.

anyway, greetz, Uli

[LigH]

Because ffvfw is able to handle WMV9, among others it expects also the WMV9 codecs being installed; if not, VirtualDub reports missing DLLs (first MSVCRTD.DLL, then WMVCORE.DLL). This might be due to static linking of the DLLs.

http://freespace.virgin.net/nick.brandon/msvcrtd/ - here you can find the MSVCRTD.DLL, unzip it into the system or VDub folder. Size: 145 KB (ZIP).

http://nic.dnsalias.com/WMNicEnc/wmfdist.exe - an early, but still official redistribution, which can be installed even without Windows Media Player 7 (the current WM9 codec pack requires that). Size: almost 4 MB.

As usual, searching for files with http://www.filemirrors.com can help, too.

[Nic]

@LigH: I updated it a while back, its the latest from the WM9 SDK :)
(strange the msvcrtd.lib are needed, thats a MFC Debug library, WMVCore.lib is needed whenever you work with the WMF SDK)

-Nic

[athos]

Originally posted by Nic
(strange the msvcrtd.lib are needed, thats a MFC Debug library [...])Perhaps I accidentily compiled some part as a debug release, will double check this for next release.

Originally posted by LigH
As usual, searching for files with http://www.filemirrors.com can help, too.
Another useful place is http://www.dll-files.com/. However the aim is that no additional files should be needed when installing ffvfw or ffdshow.

[lamer_de]

The 2003/5/2 build crashes my Vdubmod 1.4.3 without any notice instantanously after I start encoding regardless the codec and settings.

I don't rember installing WMV9, but I don't want to if I don't have to (still using wmp 6.4) and I don't get any error messages about missing dlls either. ffdshow from 2003/04/24 is installed.

The 2003/4/15 build of ffvfw works flawless though. Any ideas what I'm doing wrong?

TIA,
lamer_de

[LigH]

Originally posted by athos
Another useful place is http://www.dll-files.com/.

I would not recommend to download single DLL files: You'll never know which DLLs are required in a chain or network of DLL dependencies, or which require a registration (using regsvr32). In general I'd say, better know which package installs the complete set of DLLs (but unfortunately, some are installed only together with very disturbing but unnecessary libraries or programs - so it sometimes can indeed be a preferable way to download DLLs one by one).
_

By the way: I once tried to transcode the "Matrix Reloaded Final Trailer" (QT6, Sorenson 3 Pro, 1000x540 pixels) because even a P4 @ 1800 MHz was not able to play it smoothly; I used the QuickTime VFAPI Plugin via the AviSynth VFAPI import snippet (where I had to pull down the frame rate by 25 - ever seen a movie at 600 fps?). And due to my "cockiness", I saved it as Ogg Theora video, using ffvfw. The result looked quite strange, as if the chroma components were shifted against the luma, when I didn't crop or resize the video. After resizing down to a width of 800 pixels, it looked well, though.

Maybe not the highest priority, but you might want to watch that.

[NiTroGen]

I've moved my ffvfw archive to a new and very fast server.
Update your bookmarks --> http://users.uoi.gr/ch02029/ffvfw/

[LigH]

Here two URLs for pictures to display the Theora bug:

http://www.ligh.de/pics/Theora_1000x540.png
http://www.ligh.de/pics/Theora_800x432.png

[sh0dan]

We are getting reports of people having big problems with the ffvfw avisynth edition - I cannot reproduce.

Does anyone have any idea why this is happening:

From this thread (http://forum.doom9.org/showthread.php?s=&threadid=54633).

[Colyptic]

Anyone know if trellis quants are MPEG4 compliant?

Only reason I ask is becuase I plan on getting a hardware MPEG4 player in the next few months, but dont have access to one yet. I'd like to take advantage of the trellis quants (tested and like the results compared to without) but if its going to be a compatibility issue I will stick with what I've been doing and not use them.

[slavickas]

Originally posted by Colyptic
Anyone know if trellis quants are MPEG4 compliant?

Only reason I ask is becuase I plan on getting a hardware MPEG4 player in the next few months, but dont have access to one yet. I'd like to take advantage of the trellis quants (tested and like the results compared to without) but if its going to be a compatibility issue I will stick with what I've been doing and not use them.

trellis quants just means that quantization is done in that way that it should lower bitrate, it isn't something special like bframes, qpel or gmc

[Animaniac]

Originally posted by LigH
Here two URLs for pictures to display the Theora bug:

http://www.ligh.de/pics/Theora_1000x540.png
http://www.ligh.de/pics/Theora_800x432.png

I can confirm the same bug using ffvfw (AviSynth "lite" edition). I was transcoding a 5-minute DVD clip.

Settings:
1-pass encoding
Max I-frame interval: 240
Min I-frame interval: 1
High Quality enabled

[S_O]

When does the next build come out, because I yust read something in the CVS of ffmpeg (in the description of the file "ffv1.c"):
FFV1 codec (our very simple lossless intra only codec, compresses much better than huffyuv)If it´s as fast as Huffyuv, that would be really great! Is it possible to add this codec to ffvfw (and ffdshow) soon? That would really useful for capturing.

[Colyptic]

thank you very much Slavickas

[superdump]

I keep getting this error when I try to save my video as an avi in virtualdub (virtualdubmod just bails out and so does avs2avi):

An out-of-bounds memory access (access violation) occurred in module 'ffvfw'.

I have the latest build by athos of ffvfw 02-05-2003 and I've tried both avisynth 2.5.2/2.0.8 and mpeg2dec3.dll/mpegdecoder.dll (both nic's). What am I doing wrong? :)

Thanks

EDIT: typo

[superdump]

I just tried with the very latest vdubmod CVS and had the same problem. This is what I get from the Advanced thingy...

Sorry for the length :S I will edit this lot out when my problem has been solved. I hope this helps. Thanks.

EDIT: Removed report. If any developers want it then just say and i'll PM you with it or something. Thanks for the silence people. :p :)

[superdump]

I've just added the line "YV12toYUY2(interlaced=false)" in my avs and unchecked "force input to YV12" and selected full processing mode in vdubmod and it's working..... So I'm guessing it's some sort of problem with YV12.

Thanks

EDIT: well at least it started this time....

An out-of-bounds memory access (access violation) occurred in module 'libavcodec_enc'...
...while compressing frame 3446 from 07366000 to 07400020 (VideoSequenceCompressor.cpp:373)...
...while running thread "Processing" (thread.cpp:90).

EDIT 2: ok i changed the motion estimation things from SSE to SATD and it works now, but it still doesn't like YV12 unless I use YUY2 in the avs but it doesn't mind fast recompress in vdubmod. I dunno. Crazy. :)

EDIT 3: This is very strange. I tried removing the YV12toYUY2 conversion and left the motion estimation options as SATD. I had tried this previously but it didn't work, it works now. Without b-frames at my kind of bitrate ffvfw is too blocky during motion.

[Wilbert]

@Milan,

Could you have a look at this thread:
http://forum.doom9.org/showthread.php?s=&threadid=52503

Summarized:
"The install app for AVIsynth v2.52 wipes out a setting in your VFAPI config, and the result of this is no VFAPI "writing" support."

Somehow the installation of the included ffvfw overwrites some VFAPI settings ...

[superdump]

*bump*

Has/is anything new and significant been/being added to ffvfw?

[Animaniac]

The ffvfw and ffdshow CVS trees are pretty active (changes have been made >=29 hours ago)... I'd really appreciate new builds of each. :D Milan, what's ff_libmad?

[Dark-Cracker]

yes yes yes a new build with MPEG matrix for divx 3 :) plz plz plz