• MPlayer
• MEncoder
• 7-Zip sfx

This build corrects serious issues with 27200c-f29b3d9. I’ll include the changes from that build here and remove its post since it’s just broken everywhere but on my computer ._.

27200c-f29b3d9 changes:
All dependencies were recompiled, this build fixes the 2GB file problem.

New Dependencies:

Short Log:

• libass
  • Fixed ascender/descender calculations for empty lines.
  • Fixed vertical centering to match VSFilter’s behavior.
  • Fixed borders when one of the x/y components is 0.
  • Fixed animations with equal start and end time.
  • The \clip rectangle is now limited to the screen area.
  • Fixed reading of Bold/Italic values off Style lines.
  • Fixed several memory leaks.
  • Vector \clips that fail rendering now don’t crash libass.
  • Assume 4:3 AR when calculating missing PlayRes to mimic VSFilter behavior.
  • Fixed timings in \k and \ko tags.
  • Add extra space after an italic/non-italic change to compensate for the letter inclination.
  • Implement BorderStyle 3 (opaque box).
  • Support the \q tag
  • Fixed the colision detection of lines to consider their bounding box instead of only the vertical position.
  • Add glyph emboldening fallback for the case there is no bold variant of the requested font.
  • Use light autohinting by default in libass
• MPlayer
  • New debanding filter, -vf gradfun.
  • Changed libamr support to libopencore-amr.
  • Fixed a bug on the channel reordering code for 24bit audio, it caused broken output.
  • -vo yuv4mpeg now has disabled RGB support. (I didn’t even know it could do that)
  • -vf halfpack,pallete,rgb2bgr,yuy2 have been disabled.
  • -vf pp no longer supports the “hex mode”.
  • Added -vo matrixview.
  • Enable colored console messages on the default config.
• FFmpeg
  • Theora decoding fixes.
  • FLV demuxing fixes.
  • Added the AAC encoder. It is still considered experimental.
• x264
  • Added aq-mode=2. This significantly improves SSIM.
  • SSSE3 code improvement for avg2_w16, 1.5% faster overall on Core2
  • Added QPRD as subme=10.
  • Fixed a nondeterminism with threads and subme>7
  • Faster bidir_rd plus some bugfixes
  • 15% faster refine_bidir_satd, 10% faster refine_bidir_rd (or less with trellis=2)
  • Various 1-pass VBV tweaks, up to 1db better PSNR.
  • Macroblock-tree ratecontrol added. Disables b-pyramid if enabled (default).

Full Logs

• MPlayer
• FFmpeg-mt
• libass
• x264

• MPlayer
• MEncoder
• 7-Zip sfx

Now uses a new libass.

libass now supports most of the new VSFilter tags, \p and vector \clip()s and \iclip()s.

Built with x264 r1183.

MD5Sums:
97046643452cabd90c2ad47bb9b4fd40 *mplayer.exe
891e4d5435cee944b4f9f972b07fd1f9 *mencoder.exe
64d666f364ba2599b9a5663212d3bdcf *mplayer-20090719-e7dab44-f2086b0.7z.exe


libass Shortlog:

• Support for \p and vector \clip() and \iclip().
• Support for \iclip, \xbord, \ybord, \fax, \fay, \xshad and \yshad, all animatable.
• Support for \u and \s.
• Optimizations for \be and \blur.
• Support for using \be and \blur on the same line; \be is applied first.
• \blur is now animatable and doesn’t make artifacts.
• Support for \t using 0 for the end time.
• libass now has 1/8th pixel precision, like VSFilter.
• Tweak font height metrics to better match Windows GDI.
• Fix crash when Format: line isn’t specified.
• libass now isn’t limited by MAX_GLYPHS and MAX_LINES.
• libass now properly renders blank lines (\N\N).
• Hash calculation (for internal structs) bugfix.
• Memory leaks plugged.
• libass now does a naive PAR correction for native resolution renderers (EOSD).

MPlayer Shortlog:

• Don’t print the FAAD assuming 128kbit/s message. It doesn’t affect playback.
• Now uses the standalone libass from git://greg.geekmind.org/libass.git.
• Better support for inaccurate chapter boundaries.
• Frontends can now pass bigger paths/URLs to MPlayer in slave mode.
• Add more FourCC/TwoCCs.
• Now prints chapter information when run with -identify, for frontends.

FFmpeg-mt Shortlog:

• H.264 minor speedups and fixes.
• Frame threading support for VP3.
• New MP3 parser.

x264 Shortlog:

• Better sub-p8×8 mode decision.
• Better heuristics for early termination on chroma encoding, x264 is at least 1% faster.
• Faster lossless encoding.
• SSSE3 chroma MC. ~70% faster chroma MC on 32-bit Conroe.
• Improved SSSE3 intra_sad_8×8c.

Full Logs:

• libass
• MPlayer
• FFmpeg-mt
• x264

I have the patch to fix it already, but I won’t be able to release a new build today; probably tomorrow or the day after.