Sunday, December 08, 2013

lossless encoding from H264 to MPEGTS

For editing purposes, I needed a good way to losslessly encode video from my Canon 5D Mark II's H264 source to a format that allows me to concatenate multiple video files into one large one.  Only a few video formats have the ability to concatenate multiple files together: mainly mpg and mpegts.  FFmpeg is always getting better (thank you, guys!), and has a nice formula for doing this here:,%20merge)%20media%20files

The formula relies on a specification from the H264 standard known as AnnexB.  Here are a couple articles about Annex B, though both are above my head:

The H.264 spec itself:

My goal was to losslessly extract the video and audio from the individual videos I shot using the Canon and concatenate them all into one long-length video.

Here's what the Canon video file container looked like to FFprobe:
ffprobe $VIDEO 2>&1 | grep Stream
    Stream #0:0(eng): Video: h264 (Constrained Baseline) (avc1 / 0x31637661), yuvj420p(pc, smpte170m), 1920x1080, 45694 kb/s, 29.97 fps, 29.97 tbr, 30k tbn, 60k tbc (default)
    Stream #0:1(eng): Audio: pcm_s16le (sowt / 0x74776F73), 48000 Hz, stereo, s16, 1536 kb/s (default)

Note the video stream is H264 and the audio stream is waveform PCM.

Extract the Video
I did this with an FFmpeg command line:
ffmpeg -i $VIDEO -c copy -bsf:v h264_mp4toannexb -f mpegts $NEWVIDEO

Note that you'll need the latest and greatest FFmpeg.  Using an older version of FFmpeg from RPM Fusion's repo )FFmpeg 1.0.8), I got a bunch of these errors when I tried to concatenate individual MPEGTS files.
[aac @ 0x6727e0] channel element 1.10 is not allocated
[aac @ 0x6727e0] More than one AAC RDB per ADTS frame is not implemented. Update your FFmpeg version to the newest one from Git. If the problem still occurs, it means that your file has a feature which has not been implemented.
[aac @ 0x6727e0] channel element 3.15 is not allocated
[aac @ 0x6727e0] Error decoding AAC frame header.
[aac @ 0x6727e0] Reserved bit set.
[aac @ 0x6727e0] channel element 1.8 is not allocated
[aac @ 0x6727e0] channel element 2.0 is not allocated
[aac @ 0x6727e0] Sample rate index in program config element does not match the sample rate index configured by the container.
[aac @ 0x6727e0] Reserved bit set.
    Last message repeated 1 times
[aac @ 0x6727e0] Number of bands (4) exceeds limit (2).
[aac @ 0x6727e0] SSR not implemented. Update your FFmpeg version to the newest one from Git. If the problem still occurs, it means that your file has a feature which has not been implemented.
[mpegts @ 0x676640] decoding for stream 1 failed
[mpegts @ 0x676640] Could not find codec parameters for stream 1 (Audio: aac ([6][0][0][0] / 0x0006), 8 channels (FL+FR+FC+LFE+BL+BR+FLC+FRC), s16, 548 kb/s): unspecified sample rate
Consider increasing the value for the 'analyzeduration' and 'probesize' options
Input #0, mpegts, from 'concat:test.ts|test2.ts':
  Duration: 00:03:14.17, start: 1.400000, bitrate: 131495 kb/s
  Program 1 
      service_name    : Service01
      service_provider: FFmpeg
    Stream #0:0[0x100]: Video: h264 (Constrained Baseline) ([27][0][0][0] / 0x001B), yuvj420p, 1920x1088, 29.97 tbr, 90k tbn, 180k tbc
    Stream #0:1[0x101](eng): Audio: aac ([6][0][0][0] / 0x0006), 8 channels (FL+FR+FC+LFE+BL+BR+FLC+FRC), s16, 548 kb/s
[mpegts @ 0x672c20] sample rate not set

Once I downloaded the latest FFmpeg from FFmpeg's git repo

the errors disappeared.  Hooray for updated code!

Extract the Audio
My second task was to yank the PCM audio track from the Canon and save that off to a file for concatenation later on.  I did so using this command to extract only the audio stream (-c:a) from the source video file the Canon outputs:
ffmpeg -i $VIDEO -c:a copy $NEWAUDIO

Concatenate the Individual Video Files
Next, I combined the extracted video streams that were saved to MPEGTS format into one large file.  That command looked like this:
ffmpeg -i "concat:file1.ts|file2.ts|file3.ts" -c copy $FINAL.ts

Concatenate the Individual Audio Files
I did the same for the audio:
ffmpeg -i "concat:file1.wav|file2.wav|file3.wav" -c copy $FINAL.wav

Edit to Taste
Knowing my goal was to have as close as lossless video and audio, I did some audio sweetening in Audacity and reencoded the $FINAL.wav as an MP2 audio file.  This is lossy, but
1) the MPEGTS container prefers MPEG audio codecs.
2) FFmpeg doesn't yet have an encoder for MPEG-4 ALS lossless just yet.

Combine the Long-length Video and Audio Streams
Lastly, I combined both lossless video and lossy audio streams using this command into a:
ffmpeg -i $FINAL.mp2 -i $FINAL.ts -c:a copy -c:v copy -map 0:0 -map 0:1 $COMBINED.ts

The -map feature in FFmpeg is the best thing since sliced bread:

You can take an audio or video stream from one file and combine it with an audio or video stream from another file to save into a new file.  This assumes the container format supports those codecs.  Sweet!

Sync Issue Averted
I noticed that if I combined the individual audio and video streams into files that were greater than an hour and a half long, FFmpeg would introduce audio-visual sync issues towards the end of that long duration video.  My solution was to combine the audio and video streams into smaller transport stream files first and then combine those smaller, complete files into one big file.  This way, I had no a/v sync issue.

Anyway, that's my latest adventure into the world of lossless video editing using my Canon and FFmpeg.