• MPlayer
• MEncoder
• 7-Zip sfx (not yet)

New Years build. Sorry for the (long) delay.
I know it’s the 12th already, but it’s still the 11th in my local time :)
EDIT: Uploaded a fixed build with working automatic MPEG-2 threaded decoding.

Short Log:

• libass:
  • Use a locale-independent strtod, prevents parsing errors on locales where , is used as decimal separator.
  • Fix a bug where inverse clips would not be cleared, eventually rendering subtitles invisible
  • Improved rotation cache accuracy
  • Now supports scripts up to 50MB, up from 10MB
  • Do not generate border if a glyph is collapsed (no width or no height)
• MPlayer:
  • MPlayer now prevents the Windows screensaver from running while playing video. This can’t be disabled for now at least.
  • MPlayer now supports using FFmpeg’s stream implementations. Current syntax is ffmpeg://URL, for example ffmpeg://http://example.com/file.ogg
  • Added IJLV, MVJP, HDYC, V422, YUNV, UYNV, UYNY, uyv1, 2Vu1, P422, VYUY, Y42B, eatgv, yuvs, slif, AVI1, AVI2, MTSJ, R210, zjpeg, AV1x, AVup, VDTZ, uldx and vspx FourCCs
  • Added audio format 0×55005354 to mp3
  • Add support for “Forward Uncompressed” decoding via FFmpeg
  • Find and take into account stream start time for ogg files (Fixes #1585)
  • Matroska Improvements:
    • Now uses a new EBML parser that should allow more improvements to the demuxer.
    • Now supports compressed subtitle streams.
    • Load FLAC metadata in the proper field, fixes possible conflicts when seeking / switching streams.
    • Make time arithmetic more precise in one more case
    • Fix ordered chapter timeline building memory corruption, caused a crash when a file never referenced itself, only segments from other files.
    • Fix ordered chapter timeline if playback starts in the middle of the stream.
    • Detect and warn about virtual timeline nesting (ChapterSegmentEditionUID), which is not yet supported.
    • Support for multiple editions. Added option “-edition” for explictly choosing an edition. Prefers the edition marked as default if there is one, or the first edition.
    • -identify output: use ordered chapters length for ID_LENGTH
    • Respect -forceidx.
    • Improved seeking with generated index (-idx / -forceidx).
  • Improved teletext support, support for teletext on lavf demuxers. Added support for DVB teletext.
  • Support ISDB-Tb tunning in Brazil (SBTVD)
  • Better av_log handling (now the handler is always installed, not only when using a ffmpeg decoder).
  • Fix channel layout for ffvorbis and ffaac
  • Make dvdnav also print info about audio streams with unknown language, just like dvd://
  • dvdnav: print ID_SID_…_LANG, just like dvd://
  • Fixed the dvdnav < -> sid mapping
  • Implement -dvd-speed support for dvdnav.
  • Support the range syntax (like dvd://2-5) also for dvdnav.
  • Prevent dvdnav from hanging when reaching the last frame of a video stream.
  • Remember which subtitle was auto-selected for dvdnav due to -slang. This makes -slang work when used with menu navigation or in general when the SID changes during playback.
  • Disable audio when initializing the filter chain fails, prevents a crash.
  • Remove pointless and broken -tsprog handling from lavf demuxer, mplayer already calls IDENTIFY_PROGRAM with the right arguments and that actually works, unlike -tsprog.
  • Add support for 8 channel audio. Where 8 channel support is non-trivial (e.g. ao_dsound), at least fail gracefully.
  • Calculate length of ASF files with more than 1 second precision.
  • Allow lavf demuxer to also probe and play files < 32 kB (full probe buffer size).
  • Try to detect broken files with unaligned chunks and hopefully play them.
  • Keep aspect by adding black bars in window mode too if necessary
  • Support 16-bit-per-component RGB formats (rgb48[bln]e) in option parser.
  • Allow vf_scale to output to RGB48, though still prefer the 24 and 32 bit formats.
  • Add support for decoder reordering of pts values, using ffmpeg’s reordered_opaque feature if it seems useful. Added -pts-association-mode option to force a particular mode.
  • Make fribidi (i.e. right-to-left) support work with embedded subtitles.
  • Update OSD information after filtering, gives more accurate timestamps when not using vf_expand for drawing OSD. This makes vf_expand output information from the previous frame though.
  • When using -v, print also which lavc codec has been used, not just that the init was ok.
  • Fix mangling of 24-bit audio during channel reorder.
  • Fix a crash when fontconfig is unable to select a font.
  • Fix a case when the OSD would render a non-ass subtitle line even when libass is being used, causing two lines to appear.
  • Added ffv210, Aura and Aura 2 decoders.
  • Added cdg to lavf preferred list.
  • OpenGL VOs now support all planar YUV formats.
  • Make YUV->RGB conversion coefficients selectable for -vo gl.
  • Add support for 440p colorspace.
  • Add support for 16-bit per component YUV formats.
  • Add support for YUV with alpha.
  • FFmpeg’s AC3 and DTS decoders are now officially the default decoders.
  • Added support for either-endian AC3/SPDIF and convert as necessary.
  • Added support for AC3 with -demuxer rawaudio.
  • Support decoding of 8-bit-per-component dpx files.
  • Add support for JPEG2000 via FFmpeg/OpenJPEG.
  • MatrixView VO is now upstream.
  • OpenGL VOs now support PC< ->TV levels conversion through the levelconv parameter.
  • Improved EOF handling on seeks.
  • Support for files >2GB on HTTP and FTP streamers.
  • Support more formats for -ao pcm in raw mode.
  • Add support for parsing MLP and TrueHD.
  • Stop trying mpegpes first for mpeg-1/2 files. Such setup is not very common and produces misleading error messages.
• FFmpeg:
  • Stricter TMV and H.263 probes, prevents false positives
  • Error when VC-1 interlaced frames are found
  • Several theora fixes
  • Allow autodetection of E-AC3 and demuxing it from MPEG-TS
  • Detect Windows Media DRM and warn if no key was provided
  • Support for Intel and Midvid JPEG video codecs
  • Added Speex support to Flv and Ogg muxers
  • Skip padding correctly on ID3v2 tags
  • Support for reading WavPack and TTA tags
  • Support for demuxing teletext in DVB streams
  • Added support for some LATM AAC streams on MPEG-TS, works with MPlayer
  • Support ALS and MPEG-4 Lossless
  • Support for Flash screen codec v2
  • Support for several iTunes metadata on mp4 / mov
  • WMA fixes
  • Several APE optimizations
  • Some VP3/Theora optimizations
  • Use LZW compression in gif encoder
  • Support Weight-P and MB-tree on libx264 wrapper
  • Fix PTS on OGM
  • Several metadata parsing improvements
  • Added CD+G demuxer and decoder
  • Added Auravision Aura and Aura2 decoders
  • Added Deluxe Paint Animation demuxer and decoder
  • Support raw 1-bit video in MOV
  • libavcodec now export fullrange flag and color information for H.264
  • Support for Dirac in Ogg
  • Snow got a commit! Whee!
    • Performance improvement on a wavelet function
  • AAC decoding speedups
  • Implement alpha decoding for BGR HuffYUV
  • SIPR16k decoder
  • Several H.264 optimizations, involving rearchitectures involving stitched up gooses
  • Added support for Vorbis 6.1 and 7.1 configurations.
  • Added Bink demuxer and audio decoder
• x264:
  • Weighted P-Frames prediction support. Vastly improved fades, breaks decoding on AppleTV (not fixed, most likely won’t), CoreAVC 1.x (fixed on 2.0) and Flash 10 (fixed on betas).
  • Fix memory leaks and reduced memory usage
  • Made b-pyramid spec compliant, added support for strict b-pyramid (required for BDs)
  • Improved initial QP selection on CRF mode
  • Pyramid + VBV fixes
  • Remove non-mod16 warning, not being mod16 has no practical noticeable effects and only caused confusion
  • CABAC, CAVLC, motion compensation, bime, qpel-RD optimizations
  • MB-tree fixes, MB-tree now supports B-pyramid
  • Code size reductions, improving cache efficiency
  • Fixed extremely rare infinite loop in 2-pass VBV
  • All boolean options now are only written to the SEI as 0 or 1
  • Support for slice threading
  • Made subme=1 even worse, but faster
  • Fixed bugs in 2-pass ratecontrol, most relevant with many threads and large I-frames (like anime).
  • Support for Periodic Intra Refresh
  • Faster me=tesa, direct=auto
  • Fix the oldest-ever bug in x264: b16×8 analysis used the wrong width for predict_mv.

Full Logs

• MPlayer (plain text)
• FFmpeg-mt (plain text)
• libass (plain text)
• x264 (plain text)

• MPlayer
• MEncoder
• 7-Zip sfx

Build (very) late due to my computer borking :(
Rebuilt everything on MSYS. Updated theora to r16629. Built with fribidi.

Short Log:

• libass:
  • Emulate VSFilter’s retarded behavior on opaque box widths.
  • Emulate VSFilter’s shearing behavior.
  • Scale kerning and horizontal letter spacing with ScaleX.
  • Fix glyph overlap blending.
  • Fix crash when rendering empty, illegal drawings.
  • Calculate freetype-italicized glyph width correctly.
  • Revert “Merge glyph and outline bitmap better”. Looked better in some situations, worse in lots of others.
  • VSFilter stupidity compatibility hack: Treat \a8 as \a5.
  • Disable kerning by default and make it configurable.
    • Kerning causes problems with many (broken) fonts. VSFilter doesn’t use kerning so these problems won’t be appearant, while libass used kerning by default.
    • For compatibility reasons, disable kerning by default. In addition, make it configurable through style overrides and the Script Info header.
    • Put “Kerning: yes” into the Script Info header to enable kerning. With MPlayer, using “-ass-force-style Kerning=yes” will always enable kerning. That is equivalent to “ass-force-style=Kerning=yes” in the config file.
  • Render NBSP as a regular space. Fixes fonts that don’t have the “Non-breaking space” glyph.
  • Optimize shadow bitmap creation.
  • Fix alpha calculation in some cases. Fixes rendering on XBMC and MPlayer with -vo direct3d.
• MPlayer:
  • Disable threading on MPEG-2 / MPEG-1 video, breaks interlaced video decoding.
  • Support for TrueHD in Matroska and in transport streams (Blu-Rays)
  • Support for LPCM on Blu-Ray.
  • Fix MPEG-4 ASP, H.264, MPEG-2, AAC, MP2, DTS, AC3 and E-AC3 on transport streams with -demuxer lavf.
  • Fixed incorrect channel ordering for libavcodec audio codecs, namely ffac3, ffdca, ffflac, ffaac and fftruehd.
  • Changed default AC3 decoder from liba52 to ffac3. The only reason the former was used was because ffac3 gave wrong channel order.
  • Use the new metadata API on -demuxer lavf. On MP3 this can decode UTF-16 tags and prints more tags on the output.
  • Changed default MP3 demuxer to libavformat. MPlayer’s default demuxer reports wrong length / bitrate for VBR files and now libavformat is also capable of reporting tags.
  • Support for WMAPro audio.
  • Support for DPX images for input.
  • Support for 48-bit PNG.
  • Support for TwinVQ audio and vqf files.
  • Support for SGI images.
  • Fix demuxing if Wave64 files.
  • Handle demuxers that don’t report position by checking the byte position on the file. The mf:// demuxer now reports position since the fallback is not usable in this case.
  • Added support for 32-bit audio samples.
  • Fixed decoding of some MOV files.
  • Support -1 and -2 values for -aid, -vid and -sid (meaning “auto” and “disabled” respectively). Add a -nosub option that is equivalent to -sid -2.
  • Support for subtitles on audio-only files (for lyrics, etc). They are printed to the console as they come.
  • Support for DEP (Data Execution Prevention) on Windows.
  • Linux stuff you wish was available for windows: vastly improved VDPAU video output, replaced the previously default XV as the default video output (uau repository only).
• MEncoder:
  • Updated x264 encoder to conform with x264 r1260 API.
  • Make sure all delayed frames are written on FLUSH_FRAMES when using x264.
• libswscale:
  • MMX YV12 to BGR32 converter.
  • YUV->YV16 scaler.
• FFmpeg:
  • Updated ASM code shared with x264 and other NASM code to support Win64. Maybe I should try making a 64bit build again…
  • Support for Sony Wave64 files. They’re basically regular Wave files with large file support.
  • Fix a crash on SVQ1.
  • Support for ID3 and APE metadata on Musepack SV7. APE metadata parsing bugfixes.
  • TwinVQ decoder.
  • Fixed decoding of video from MTV files.
  • Support for Blu-ray subtitles, this can’t be used by MPlayer for now, at least.
  • Support for LPCM on Blu-Ray.
  • 5% faster vorbis decoding on Core2 because pengvado got a good roll on GCC’s random code generator.
  • Shut up useless warnings on ffmp3.
  • Several false positive fixes on demuxer probes.
  • Set ID3v1 default genre to unknown instead of Blues.
  • Core Audio Format demuxer.
  • Print a warning if a format has been detected by random guessing (low score).
  • Allow encoding frames taller than 2800 pixels.
  • MD Studio demuxer and Atract1 decoder.
  • Several security-related patches integrated from Chrome.
  • Support the new x264 api.
  • Support UTF-16 in ID3v2 tags by converting them to UTF-8.
  • Export mpv/mp4 major and compatible brands as metadata.
  • TIFF decoding fixes.
  • AAC decoder now supports spec-compliant channel layout configuration.
  • Support writing ID3v2 metadata to mp3 files.
  • Support for contant quantizer encoding with libtheora.
  • Convert an AC-3 error to a warning to allow certain broken files that happen to work when not aborted.
  • Support for uncompressed TARGA images.
  • Support Vorbis channel layout according to the Vorbis I specification.
  • Allow combined TrueHD/AC3 tracks on BDs to be demuxed as two separate streams.
• x264:
  • Several 1-pass and 2-pass VBV fixes. Support for single-frame VBV.
  • Use -mfmpath=sse by default on compilation. Faster floating point calculations in C code.
  • Optimize x264_exp2fix8, 22->13 cycles on Core2.
  • Faster mbtree propagate and less memory usage.
  • Fix bug in b16×16 qpel RD where an incorrect cost was used.
  • Fixed an iron age bug (r369) with intra-analysis on B-frames.
  • Multi-slice encoding support. Slicing support is available through three methods (which can be mixed):
    • --slices: sets a number of slices per frame and ensures rectangular slices (required for Blu-ray). Overridden by either of the following options:
    • --slice-max-mbs: sets a maximum number of macroblocks per slice.
    • --slice-max-size: sets a maximum slice size, in bytes (includes NAL overhead).
  • Implement macroblock re-encoding support to allow highly accurate slice size limitation.
  • Faster me=tesa.
  • Threaded lookahead, reducing the amount that lookahead bottlenecks encoding specially with lookahead intensive settings. Can be tuned with --sync-lookahead, default being threads+bframes. This buffer is unrelated to --rc-lookahead.
  • Fix RD early skip, small quality improvement and speedup.
  • x264’s VBV now works on a NAL level, taking into acount escape codes.
  • Reorder reference frames optimally on second pass. Needs both passes to have the same ref count value.
  • SSE4 version of 4×4 IDCT, 27->24 clocks on Nehalem. Just so Holger gets to use movsd.
  • Fix bug where x264 generated bitstreasm with insane SAR values.
  • Avoid scenecuts on flashes and similar situations. Works on 1-frame flashes with b-adapt 1, bframe-count length flashes with b-adapt 2. Flashes are defined as any scene which lasts a very short period of time before returning to a previous scene, such as one caused by a camera flash.
  • Fix a very rare overflow that triggered an assert failure; bug was as old as b-adapt.

Full Logs

• MPlayer
• FFmpeg-mt
• libass
• x264