Originals

Original audio files for Dharma teaching and meditation instructions at Wat Kow Tham.

• 10-20 April 2018 by Dr.Marut Damchaom

How to encode and upload to YouTube

Instructions how to record, process and upload audio lectures for people to study, based on wat-kow-tham/checklist#1

Current workflow:

.m4a (IPhone) --> .mp4 (ffmpeg) --> YouTube

Enhanced workflow:

.m4a (IPhone) --> normalize (ffmpeg) --> .opus (ffmpeg) --> .webm (ffmpeg) --> YouTube

FFmpeg tutorial

The essential information you need to know about FFmpeg:

• it consumes videos and audios (inputs)
• it produces videos, audios, text files (outputs)
• most video files contain multiple streams
  • usually there is at least one audio stream and one video stream
  • the first stream of each type is used by default

running

ffmpeg -hide_banner

By default ffmpeg produces somewhat confusing output, so -hide_banner helps it to keep straight to the point.

get information about audio/video file

$ ffmpeg -hide_banner -i 20180412_002.m4a
Input #0, mov,mp4,m4a,3gp,3g2,mj2, from '20180412_002.m4a':
  Metadata:
    major_brand     : mp42
    minor_version   : 0
    compatible_brands: isommp42
    creation_time   : 2018-04-12T08:39:53.000000Z
    com.android.version: 6.0
  Duration: 01:09:38.81, start: 0.000000, bitrate: 130 kb/s
    Stream #0:0(eng): Audio: aac (LC) (mp4a / 0x6134706D), 48000 Hz, stereo, fltp, 128 kb/s (default)
    Metadata:
      creation_time   : 2018-04-12T08:39:53.000000Z
      handler_name    : SoundHandle

Here we specify .m4a audio file as input to ffmpeg, and then ffmpeg reports that there is one input #0 with one stream #0:0, which is audio with some characteristics.

generate 1 hour white color video

We need to create video stream, because YouTube refuses to accept files with only audio inside. First we need to learn how to create blank video.

The command below uses color filter to produce video stream with white background. -t parameter limits time of the resulting video to 1 hour. Last parameter is filename of resulting file white.webm - for ffmpeg it is output #0.

$ ffmpeg -hide_banner -t 3600 -filter_complex color=white white.webm
Stream mapping:
  color -> Stream #0:0 (libvpx-vp9)
Press [q] to stop, [?] for help
[libvpx-vp9 @ 0x55a15c5de480] v1.7.0
Output #0, webm, to 'white.webm':
  Metadata:
    encoder         : Lavf58.12.100
    Stream #0:0: Video: vp9 (libvpx-vp9), yuv420p, 320x240 [SAR 1:1 DAR 4:3], q=-1--1, 200 kb/s, 25 fps, 1k tbn, 25 tbc (default)
    Metadata:
      encoder         : Lavc58.18.100 libvpx-vp9
    Side data:
      cpb: bitrate max/min/avg: 0/0/0 buffer size: 0 vbv_delay: -1
frame=   82 fps=0.0 q=0.0 size=       1kB time=00:00:02.28 bitrate=   2.2kbits/s spee
frame=  139 fps=138 q=0.0 size=       1kB time=00:00:04.56 bitrate=   1.1kbits/s spee
frame=  217 fps=144 q=0.0 size=       4kB time=00:00:07.68 bitrate=   3.8kbits/s spee
frame=  292 fps=145 q=0.0 size=       7kB time=00:00:10.68 bitrate=   5.0kbits/s spee  

Now white.webm contains 1 hour of blank white video. -filter_complex param is used to specify filters for video processing. .webm format for output was detected automatically, and then ffmpeg automatically mapped output of color filter to output stream #0:0.

create combined audio+video file from separate input streams

We now create single .mp4 video+audio file from separate .m4a stream and separate blank white video stream.

The command below tells ffmpeg to copy default input audio stream as is (-codec:a copy or -c:a copy), then produce video stream using filter (-filter_complex color=white) and stop encoding when one or another input stream ends (-shortest).

$ ffmpeg -hide_banner -i 20180412_002.m4a -codec:a copy -filter_complex color=white -shortest whiteaudio.mp4

Input #0, mov,mp4,m4a,3gp,3g2,mj2, from '20180412_002.m4a':
  Metadata:
    major_brand     : mp42
    minor_version   : 0
    compatible_brands: isommp42
    creation_time   : 2018-04-12T08:39:53.000000Z
    com.android.version: 6.0
  Duration: 01:09:38.81, start: 0.000000, bitrate: 130 kb/s
    Stream #0:0(eng): Audio: aac (LC) (mp4a / 0x6134706D), 48000 Hz, stereo, fltp, 128 kb/s (default)
    Metadata:
      creation_time   : 2018-04-12T08:39:53.000000Z
      handler_name    : SoundHandle
Stream mapping:
  color -> Stream #0:0 (libx264)
  Stream #0:0 -> #0:1 (copy)
...
Output #0, mp4, to 'whiteaudio.mp4':
  Metadata:
    major_brand     : mp42
    minor_version   : 0
    compatible_brands: isommp42
    com.android.version: 6.0
    encoder         : Lavf58.12.100
    Stream #0:0: Video: h264 (libx264) (avc1 / 0x31637661), yuv420p, 320x240 [SAR 1:1 DAR 4:3], q=-1--1, 25 fps, 12800 tbn, 25 tbc (default)
    Metadata:
      encoder         : Lavc58.18.100 libx264
    Side data:
      cpb: bitrate max/min/avg: 0/0/0 buffer size: 0 vbv_delay: -1
    Stream #0:1(eng): Audio: aac (LC) (mp4a / 0x6134706D), 48000 Hz, stereo, fltp, 128 kb/s (default)
    Metadata:
      creation_time   : 2018-04-12T08:39:53.000000Z
      handler_name    : SoundHandle
frame=  484 fps=0.0 q=28.0 size=     256kB time=00:00:17.24 bitrate= 121.7kbits/s spee
frame=  974 fps=972 q=28.0 size=     512kB time=00:00:36.84 bitrate= 113.9kbits/s spee
frame= 1913 fps=1273 q=28.0 size=    1024kB time=00:01:14.40 bitrate= 112.8kbits/s spee
frame= 2838 fps=1417 q=28.0 size=    1536kB time=00:01:51.40 bitrate= 113.0kbits/s spee
...
frame=68597 fps=2077 q=28.0 size=   43776kB time=00:45:41.76 bitrate= 130.8kbits/s spee

fps=2077 in the last line means that ffmpeg could encode 2077 "frames per second", and 25 fps in output stream description means that our blank white video plays at 25 frames per second.

speedup video encoding

For encoding to .mp4 FFmpeg provides several presets. Blank white video doesn't need any quality controls, so we can just use the fastest -preset ultrafast and drop frames per second in output video (frame rate) to 5 (-r 5).

$ REC="20180412_002"; time ffmpeg -hide_banner -i "${REC}.m4a" -c:a copy -filter_complex color=white -shortest -preset ultrafast -r 5 "${REC}.mp4"

This command uses shell variable to reuse filename for input and output. Using framerate of 5 frames per second makes vlc complain while playing, but should be okay for YouTube.

normalize audio and convert to .webm

Let's apply audio filter for dynamic audio normalization -af dynaudform. And because audio processing requires reencoding, let's save the result into open .webm format that is unencumbered by patents.

$ REC="20180412_002"; time ffmpeg -hide_banner -i "${REC}.m4a" -af dynaudnorm -filter_complex color=white -deadline realtime -shortest -r 5 "${REC}.webm"

-deadline realtime is an option to speedup VP9 encoding. ffmpeg chooses Opus audio codec and VP9 video codec for .webm format automatically.