Recherche avancée

Sur d’autres sites (9442)

• AVCodecContext settings for H264 (1080i)

  23 septembre 2013, par moose

  I'm trying to configure x264 for 1080i capturing. Most of these settings below are found in different examples. However, compiled together they don't work. ffmpeg API reports no error, but avcodec_encode_video() always returns zero.

  Some of the numbers are strange to me... for example, gop_size. Isn't 250 too high ?

  Event you can't offer the final answer, I'm still interested in any kind of comment on this subject.

  pCodecContext->codec_type              = AVMEDIA_TYPE_VIDEO;
  
  pCodecContext->codec_id                = CODEC_ID_H264;
  
  
  
  pCodecContext->coder_type              = FF_CODER_TYPE_AC;
  
  pCodecContext->flags                  |= CODEC_FLAG_LOOP_FILTER | CODEC_FLAG_INTERLACED_ME | CODEC_FLAG_INTERLACED_DCT;
  
  pCodecContext->me_cmp                 |= 1;
  
  pCodecContext->partitions             |= X264_PART_I8X8 | X264_PART_I4X4 | X264_PART_P8X8 | X264_PART_B8X8;
  
  pCodecContext->me_method               = ME_UMH;
  
  pCodecContext->me_subpel_quality       = 8;
  
  pCodecContext->me_range                = 16;
  
  
  
  pCodecContext->bit_rate                = 10 * 1024 * 1024; // 10 Mbps??
  
  pCodecContext->width                   = 1920;
  
  pCodecContext->height                  = 1080;
  
  
  
  pCodecContext->time_base.num           = 1;    // 25 fps
  
  pCodecContext->time_base.den           = 25;   // 25 fps
  
  
  
  pCodecContext->gop_size                = 250; // 250
  
  pCodecContext->keyint_min              = 25;
  
  pCodecContext->scenechange_threshold   = 40;
  
  pCodecContext->i_quant_factor          = 0.71f;
  
  pCodecContext->b_frame_strategy        = 1;
  
  pCodecContext->qcompress               = 0.6f;
  
  pCodecContext->qmin                    = 10;
  
  pCodecContext->qmax                    = 51;
  
  pCodecContext->max_qdiff               = 4;
  
  pCodecContext->max_b_frames            = 3;
  
  pCodecContext->refs                    = 4;
  
  pCodecContext->directpred              = 3;
  
  pCodecContext->trellis                 = 1;
  
  pCodecContext->flags2                 |= CODEC_FLAG2_WPRED | CODEC_FLAG2_MIXED_REFS | CODEC_FLAG2_8X8DCT | CODEC_FLAG2_FASTPSKIP;  // wpred+mixed_refs+dct8x8+fastpskip
  
  pCodecContext->weighted_p_pred         = 2; // not implemented with interlaced ??
  
  pCodecContext->crf                     = 22;
  
  
  
  pCodecContext->pix_fmt                 = PIX_FMT_YUV420P;
  
  
  
  pCodecContext->thread_count            = 0;

• How to calculate ffmpeg output file size ?

  25 septembre 2011, par poundifdef

  I am using ffmpeg to convert home videos to DVD format and want to calculate the output file size before doing the conversion.

  My input file has a bit rate of 7700 kbps and is 114 seconds long. The audio bitrate is 256 kbit (per second ?) The input file is 77MB. To get this information I ran :

  mplayer -vo null -ao null -frames 0 -identify input.MOD

  So in theory, the input file should have (roughly) a file size of :

  ((7700 / 8) * 114) / 1024

  That is, (7700 / 8) is kilobytes/second, multiplied by 114 seconds, and then converted to megabytes. This gives me 107MB, which is way beyond my 77. Thus I am skeptical of his formula.

  That said, after converting the video :

  ffmpeg -i input.MOD -y -target ntsc-dvd -sameq -aspect 4:3 output.mpg

  The numbers seem to make more sense. Bitrate is 9000 kbps, and applying the above formula, I get 125MB, and my actual output file size is 126MB.

  So, two questions :

  1. How do I factor the audio bitrate into this calculation ? Is it additive (video file size + audio file size) ?

  2. Do DVDs always have a 9000 kilobit/second rate ? Is that the definition of a DVD ? Or might that change depending on video quality of my input video ? What does "-target ntsc-dvd" guarantee about my video ?

  3. Why does my input file not "match" the calculation, but the output file does ? Is there some other variable I'm not accounting for ?

  What is the correct way to calculate filesize ?

• converting .mov file to .h264 file

  29 août 2011, par Robin Rye

  ok, this is the case, i actually want to parse frames from a mov file. get the encoded h264 frames. and i've managed to do so by using ffmpeg but when i try to make a movie again by using ffmpeg -i test* test.mov i get test00: Invalid data found when processing input so there is something not correct with the structure of the frames. as i understand it a frame should have the following appearance :

  00 00 00 01 XX data -------------

  where XX is say whether it is a I-,P- or B-frame. or more specifically type(XX) = 0x0F && XX says if it is I(type(XX) = 5 ?),P(type(XX) = 7 ?) or B(type(XX) = 8 ?) frame. I'm not sure about these number, i've been looking for it but not found good sources. so that's question number one, what number should the NALU be for the different frames ?

  anyway, when i use av_read_frame on the mov file, i get frame that look like this :

  4B = size, 1B = XX and then data. (at least this is what i think i get)

  the files where i store the frames are always size long when i look at them in a hexeditor(otherwise as well of course). and XX is always 65(ie. type(XX) = 5) in the first and then 61(ie. type(XX) = 1) for a couple of frames and then back to being 65 for one frame and so on.

  i guess that these are frames like : I P P P P P P I P P P P P P P I P P P P P P P .... however then my assumption about the type numbers for the different frame types are false, which is highly likely. (any suggestion on reading about this ? except the ISO/IEC 14496-10, i don't understand it really).

  I've tried to remove the size and append 00 00 00 01 before the XX byte and the data but without success. any tips on how i could modify the frames to be valid H264 encoded frames ?