Recherche avancée

Sur d’autres sites (8945)

• flv reencode to mp4 for iphone/ipod via ffmpeg and x264 (quality issue)

  3 octobre 2011, par zeroasterisk

  There are a lot of questions on this topic, and I've read most of them and most of the google search results I could come up with.

  When I use FFMPEG to convert a FLV to a iphone3 compatble MP4 file, it just doesn't preserver enough of the quality. Yes, I've worked the hell out of -sameq and -b and -bt settings, text just isn't readable.

  Next I tried to split the video out and process it directly, using these instructions :
  https://sites.google.com/site/linuxencoding/x264-encoding-guide

  The problem is myplayer (via ffmpeg) was not able to determine the duration of the FLV (even though the metadata was set).

  (I assume) Because of that unknown duration, when I create the MP4 file, the resulting x264 file plays through super-fast while the audio plays at the normal rate.

  user@server:/tmp# mplayer -nosound -benchmark -sws 9 -vf dsize=640:480:0,scale=0:0,expand=640:480 -vo yuv4mpeg:file=>(x264 --demuxer y4m --crf 0 --preset slow --threads auto --output output.264 - 2>x264.log) 'input.flv'
  
  MPlayer 1.0rc4-4.4.5 (C) 2000-2010 MPlayer Team
  
  mplayer: could not connect to socket
  
  mplayer: No such file or directory
  
  Failed to open LIRC support. You will not be able to use your remote control.
  
  
  
  Playing input.flv.
  
  libavformat file format detected.
  
  [flv @ 0x1202460]Estimating duration from bitrate, this may be inaccurate
  
  [lavf] stream 0: video (vp6f), -vid 0
  
  [lavf] stream 1: audio (nellymoser), -aid 0
  
  VIDEO:  [VP6F]  1680x992  0bpp  1000.000 fps   33.4 kbps ( 4.1 kbyte/s)
  
  Clip info:
  
   audiocodecid: 6
  
   audiodatarate: 86
  
   audiosamplerate: 44100
  
   audiosamplesize: 16
  
   audiosize: 6097005
  
   canSeekToEnd: true
  
   datasize: 8609138
  
   duration: 567
  
   framerate: 2
  
   hasAudio: true
  
   hasCuePoints: false
  
   hasKeyframes: true
  
   hasMetadata: true
  
   hasVideo: true
  
   height: 992
  
   lasttimestamp: 567
  
   metadatacreator: flvtool++ (Facebook, Motion project, dweatherford)
  
   stereo: false
  
   totalframes: 1043
  
   videocodecid: 4
  
   videodatarate: 33
  
   videosize: 2316256
  
   width: 1680
  
  Using (default) progressive frame mode.Opening video filter: [expand w=640 h=480]
  
  Expand: 640 x 480, -1 ; -1, osd: 0, aspect: 0.000000, round: 1
  
  Opening video filter: [scale w=0 h=0]
  
  Opening video filter: [dsize=640:480:0]
  
  ==========================================================================
  
  Opening video decoder: [ffmpeg] FFmpeg's libavcodec codec family
  
  Selected video codec: [ffvp6f] vfm: ffmpeg (FFmpeg VP6 Flash)
  
  ==========================================================================
  
  Audio: no sound
  
  Starting playback...
  
  Movie-Aspect is undefined - no prescaling applied.
  
  [swscaler @ 0x7f0c738b9620]Lanczos scaler, from yuv420p to yuv420p using MMX2
  
  VO: [yuv4mpeg] 640x480 => 641x480 Planar YV12

  I have also tried specifying FPS, but no change in results

  user@server:/tmp# mplayer -nosound -fps 25-benchmark -sws 9 -vf dsize=640:480:0,scale=0:0,expand=640:480 -vo yuv4mpeg:file=>(x264 --demuxer y4m --fps 25 --crf 0 --preset slow --threads auto --output output.264 - 2>x264.log) 'input.flv'

  Can someone tell me how to either :

  1. fix my split A/V processing/timing/duration issues ?
  2. improve the
     quality of the FFMPEG conversion of FLV to iphone3 compatible
     format ?

• Revision a142d6fc93 : Skip redundant motion search in 4x4 level rd loop This commit makes the encoder

  16 juillet 2013, par Jingning Han

  Changed Paths :
   Modify /vp9/encoder/vp9_rdopt.c

  
  
  Skip redundant motion search in 4x4 level rd loop

  This commit makes the encoder to perform motion search only once
  per reference frame type for each 4x4/4x8/8x4 block. For bus_cif
  at 2000 kbps, the runtime goes from 253812ms -> 217817ms
  (14% speed-up) for speed 0.

  Change-Id : I5f17599ccc8cfaf93ccb4f98fcb6008af6d79e92

• Revision 91365addf8 : SSE2 8x8 inverse ADST/DCT transform This commit enables SSE2 implementation of

  13 juillet 2013, par Jingning Han

  Changed Paths :
   Modify /vp9/common/vp9_rtcd_defs.sh

  
   Modify /vp9/common/x86/vp9_idct_intrin_sse2.c

  
  
  SSE2 8x8 inverse ADST/DCT transform

  This commit enables SSE2 implementation of 8x8 inverse ADST/DCT
  transform. The runtime goes from 1216 cycles -> 266 cycles.
  For bus_cif at 2000 kbps, the overall runtime reduces from
  253707ms -> 248430ms, i.e., 2% speed-up at speed 0.

  Change-Id : Ib0372e17e9162d7b11a10d653b1c8be547c878fb