FFDSHOW is a DirectShow decoding filter for decompressing DivX, XviD, WMV, MPEG-1 and MPEG-2 movies. It uses libavcodec from ffmpeg project for video decompression, postprocessing code from mplayer to enhance visual quality of low bitrate movies, and is based on original DirectShow filter from XviD, which is GPL'ed educational implementation of MPEG4 encoder.

Notes about latest versions of FFDShow MPEG-4 Video Decoder :

- FFDShow 2006-01-23 was compiled by clsid using ICL 9+GCC 3.4.5, FFDShow 2006-01-23 SSE2** was compiled by issa using ICL 9+GCC 4.0.2 while FFDShow 2005-12-21 SSE* was compiled by bob0r using GCC 4.0.2.
* Only for CPUs with support for SSE (Streaming SIMD Extensions) instructions.
** Only for CPUs with support for SSE2 (Streaming SIMD Extensions 2) instructions. Check if your CPU has support for SSE orSSE2 instructions using Cpu-Z, a small tool which shows infos about CPU.

First of all, the FFDSHOW consumes way less CPU power than original DivX codecs, even with maximum post-processing. At the same time the post-processing seems to give excellent quality results and allows more configuring than the original DivX 5 codec. It can even automaticly reduce post-processing if your CPU is getting overloaded. The noise filter adds a kind of a rasterising effect to the picture - might be useful in some cases. The alpha version also includes experimental playback for MPEG-1 and MPEG-2 files (such as VCD and SVCD movies), which works pretty damn well. Only problem is that MPEG-2 playback can't play the SVCD clips directly from the CD, but requires the video to be extracted to the HDD by using VCDGear first.

Features of FFDShow MPEG-4 Video Decoder :

- various compression methods:
ˇ MPEG 4 compatible (Xvid, DivX 4, DivX 5) using libavcodec or XviD
ˇ Divx 3 compatible, MSMPEG4v2, MSMPEG4v1
ˇ WMV1/7, WMV2/8
ˇ H263, H263+
ˇ HuffYUV in YV12 colorspace
ˇ MPEG 1
ˇ Theora (not for regular use, format isn't stabilized yet)
ˇ MPEG 2 by using mpeg2enc code from MJPEGtools project
ˇ all Windows Media Video 9 supported encoders with output to asf/wmv file or AVI
- all common encoding modes: constant bitrate, fixed quantizer, fixed quality, two pass encoding (depends on compressor features)
- for libavcodec and XviD detailed selection of motion estimation method
- minimum and maximum I frames interval
- minimum and maximum quantizer selection, quantizer type and trellis quantization selection for libavcodec
- adaptive quantization (aka. masking) for libavcodec and XviD, single coefficient elimination
- selectable compression of starting and ending credits
- two curve compression algorithms for second pass of two pass encoding - thanks to XviD developers
- second pass simulation: although not very precise, still helpful for tweaking advanced curve compression parameters
- ability to use libavcodec internal two pass code
- image preprocessing with ffdshow image filters (latest ffdshow version must be installed)
- graph during encoding: if your encoding program doesn't provide one
- B frames support: from one to eight consecutive B frames
- support for MPEG 4 quarterpel and GMC
- selectable interlaced encoding
- decompression

Changes in FFDShow MPEG-4 Video Decoder 2005-12-21 :

- fixed GCC compilation (again)
- working on better SSA support
- support for few more SSA tags
- don't crash when performance counters aren't available
- correct (?) RGB->YV12 (JPEG) coefficients,
- dropped own DLL entrypoint, using DllMain instead
- more hash_maps
- updated libavcodec
- updated French translation