Linux workstation : audio and video

When starting the playback of a video file, VLC displays its Title tag, read from the internal metadata. Turns out some metadata fields have

typos
useless / wrong information
unsupported characters rendered as squares

which is why you'd like to fix these.

VLC permits editing these fields (add file to playlist + select it + Ctrl-j + General tab), but changes made to .mkv files aren't permanent : they're lost when re-opening the file.

Tools

The mkvmerge and mkvpropedit binaries used hereafter are from the mkvtoolnix package (project homepage).

I've not spent much time with this program, but I don't get its logic : mkvpropedit can set values but has no easy method to read such values...
... which can be done with a binary named mkvmerge
mkvtoolnix also has a GUI version, but it looks rather complex, and the CLI tools are enough for my current need anyway

Get file properties

mkvmerge -J movie.mkv | jq -r '.container.properties.title'

Set file properties

mkvpropedit --set "title=New title" movie.mkv

Since

Code is worth 1000 words

let's dive into it :

youtubeVideoUrl='https://music.youtube.com/watch?v=JozAmXo2bDE'
audioFormatCode='251'; audioFile='audio.webm'
videoFormatCode='248'; videoFile='video.webm'

yt-dlp -f $audioFormatCode "$youtubeVideoUrl" -o "$audioFile"
yt-dlp -f $videoFormatCode "$youtubeVideoUrl" -o "$videoFile"

ffmpeg -i "$audioFile" -i "$videoFile" -c:a copy -c:v copy 'audio_video.webm'

Using different audio / video formats may require transcoding (which also depends on the format of the resulting video). For example, with :

251 audio (webm)
137 video (MP4)
and to get a MP4 output file

the audio stream needs to be converted into AAC and the merge command becomes : ffmpeg -i "$audioFile" -i "$videoFile" -c:a aac -c:v copy 'audio_video.mp4'

ffmpeg -i concat:"file1|file2|file3" -c copy output

WAV to MP3 :

for audioFile in *wav; do ffmpeg -i "$audioFile" -codec:a libmp3lame -b:a 256k "${audioFile%.*}.mp3"; done

WAV to FLAC (source) :

for audioFile in *wav; do ffmpeg -i "$audioFile" -af aformat=s16:44100 "${audioFile%.*}.flac"; done

This encodes to 16 bits at 44100 hertz.

General case :

The command below will :

"open" inputFile to do stuff with it
"select" data starting at startTime of inputFile
make this selection last for duration
and copy the selection into outputFile

ffmpeg -i inputFile -ss startTime -t duration -codec copy outputFile

startTime and duration are in either a number of seconds, or in the hh:mm:ss[.xxx] format.
If no duration is specified, the process will continue until the end of the input file.

Cut a few seconds from the beginning of a video :

The command above may work pretty well. But, depending on the input file itself (I got this with a .avi file), it may end in a huge series of :

Non-monotonous DTS in output stream 0:1; previous: 103475, current: 34492; changing to 103476. This may result in incorrect timestamps in the output file.

+ audio and video not in sync anymore. This has to do with the file format itself (AVI container, mkv, mpg, mp4, ...) and codecs. see also

Re-encoding the audio track (here in mp3 format) fixed it :

ffmpeg -i input.avi -ss 15.5 -c:v copy -c:a mp3 output.avi

During the tests to find the right start time (-ss value), save time by not processing the input until the end :

ffmpeg -y -i input.avi -ss 15.5 -t 30 -c:v copy -c:a mp3 output.avi

TESTING :
outputDir='./output'; mkdir -p "$outputDir"; for input in *avi; do echo "$input"; ffmpeg -y -i "$input" -ss 15.5 -t 15 -c:v copy -c:a mp3 "$outputDir/$input"; done

REAL :
outputDir='./output'; mkdir -p "$outputDir"; for input in *avi; do echo "$input"; ffmpeg -i "$input" -ss 15.5 -c:v copy -c:a mp3 "$outputDir/$input"; done

I'd like to share a .mkv movie from my PC with MiniDLNA and play it on my TV. However, my TV automatically plays the 1^st audio track (whatever language it is) and won't let me select another one.

So far, I don't know whether this is :

a limitation in the DLNA protocol itself
or a limitation / bug in its implementation on the server-side in MiniDLNA
or a limitation / bug in its implementation on the client-side in my TV (LG, but I've read similar complaints from Samsung owners)

Since I can't debug + fix the firmware of my TV, let's see what happens if I change the order of the audio tracks in the .mkv file :

List all tracks of the movie file :

ffmpeg -i input.mkv 2>&1 | grep Stream

Stream #0.0(eng): Video: h264 (High), yuv420p, 1920x1080, PAR 1:1 DAR 16:9, 23.98 fps, 1k tbn, 47.95 tbc (default)
Stream #0.1(eng): Audio: ac3, 48000 Hz, 5.1, fltp, 448 kb/s	my TV plays this audio track
Stream #0.2(fre): Audio: ac3, 48000 Hz, 5.1, fltp, 448 kb/s	I want this one played
Stream #0.3(fre): Subtitle: srt
Stream #0.4(eng): Subtitle: srt

Change the order of tracks using -map : ffmpeg can build a new multi-track output file by picking tracks from the specified input file :
- the 1^st map of the command line is for the 1^st track of the output file
- the 2^nd map of the command line is for the 2^nd track of the output file
- and so on, you get it
ffmpeg -i input.mkv -map 0:0 -map 0:2 -map 0:1 -c copy output.mkv
This specifies the tracks of output.mkv to be :
- Track 0 : track 0 of input.mkv (video)
- Track 1 : track 2 of input.mkv (french audio)
- Track 2 : track 1 of input.mkv (english audio)
-c copy means the tracks must not be re-encoded but copied as-is.

In output.mkv, we now have :

ffmpeg -i output.mkv 2>&1 | grep Stream

Stream #0.0(eng): Video: h264 (High), yuv420p, 1920x1080, PAR 1:1 DAR 16:9, 23.98 fps, 1k tbn, 47.95 tbc (default)
Stream #0.1(fre): Audio: ac3, 48000 Hz, 5.1, fltp, 448 kb/s
Stream #0.2(eng): Audio: ac3, 48000 Hz, 5.1, fltp, 448 kb/s

We've not picked the subtitles tracks !

Same as above, with ALL tracks :
ffmpeg -i input.mkv -map 0:0 -map 0:2 -map 0:1 -map 0:3 -map 0:4 -c copy output.mkv

Last words : the whole operation lasted 1.5 minutes on my PC for a 3.9GB input file. It is actually more stressful for the HDD than for the CPU (30% of a core used).

There are several methods to get an audio track (source) :

As uncompressed audio :

ffmpeg -i movie.mp4 audio.wav

As MP3 :

one file : ffmpeg -i movie.mp4 -codec:a libmp3lame audio.mp3
more files :
- for videoFile in *mp4; do ffmpeg -i "$videoFile" -codec:a libmp3lame "$videoFile".mp3; done
- outputDir='./audio'; mkdir -p "$outputDir"; for videoFile in *mp4; do ffmpeg -i "$videoFile" -codec:a libmp3lame -b:a 256k "$outputDir/${videoFile%.*}.mp3"; done

Don't forget to specify the audio bitrate (defaults to 64kbit/s) with -b:a bitrate :: ffmpeg -i movie.mp4 -codec:a libmp3lame -b:a 256k audio.mp3
Check :: Get information (codec, bitrate, ...) about a media file

Only a snippet, as MP3 :

ffmpeg -y -ss 00:14:30 -i movie.avi -t 4.5 -codec:a libmp3lame audio.mp3

As Ogg Vorbis :

one file : ffmpeg -i movie.mp4 -codec:a libvorbis audio.ogg
more files : outputDir='./ogg'; mkdir -p "$outputDir"; for sourceFile in *mp4; do ffmpeg -i "$sourceFile" -codec:a libvorbis "$outputDir/${sourceFile%.*}.ogg"; done

For a reason unknown so far, this command _sometimes_ re-encodes the specified video file into "Ogg Theora" video format instead of extracting the audio track + converting it to Ogg Vorbis. I must have missed something important

Copy the audio track as-is with the map method :

This method should be preferred whenever possible since it only extracts the audio data with no reconversion :

preserves the original audio quality
doesn't produce an artificial "high quality" audio file from a lower quality source (which is what websites offering to download music from YouTube as MP3 do)
best possible audio quality vs file size ratio

identify the audio track :

Stream #0:0(und): Video: h264 (Main) (avc1 / 0x31637661), yuv420p, 1280x720 [SAR 1:1 DAR 16:9], 389 kb/s, 29.97 fps, 29.97 tbr, 90k tbn, 59.94 tbc (default)
Stream #0:1(und): Audio: aac (LC) (mp4a / 0x6134706D), 44100 Hz, stereo, fltp, 125 kb/s (default)

copy it into a new file :
ffmpeg -i movie.mp4 -map 0:1 -c copy audio.mp4

you can make sure this file as only 1 audio track :

Stream #0:0(und): Audio: aac (LC) (mp4a / 0x6134706D), 44100 Hz, stereo, fltp, 125 kb/s (default)

Rename the audio file based on the type of audio format :
mv audio.{mp4,aac}

Transcode files (details) :

From any format to MPEG 1 :

ffmpeg -i input.mp4 output.mpg

From any format to MPEG 2 :

ffmpeg -i input.mp4 -f dvd output.mpg

From any format to H.264 inside an MP4 container :

ffmpeg -i input.mp4 -strict experimental -vcodec libx264 -deinterlace -threads 0 output.mp4

-strict experimental was suggested by the software :

encoder 'aac' is experimental and might produce bad results.
Add '-strict experimental' if you want to use it.

From any format, convert video to H.264 + leave audio as-is, and save the result into an MP4 container :

ffmpeg -i input.mp4 -threads auto -c:a copy -vcodec libx264 output.mp4

Rotate a video :

Please note :

Videos shot with a smartphone have a Rotation metadata parameter. It is used by the built-in player to orient the video properly (source).
Looks like VLC ignores or misinterprets this parameter (?)
Despite all my efforts, I've not been able to rotate a video made with an Android phone with ffmpeg. Neither did others.
OpenShot did it very well with a few clicks.
Converting videos can be HARD on CPU, cause overheating (?) and make the system hang To workaround this, considering adjusting manually the CPU frequency.

90° :

https://stackoverflow.com/questions/3937387/rotating-videos-with-ffmpeg/9570992#9570992
	ffmpeg -i in.mov -vf "transpose=1" out.mov

For the transpose parameter you can pass:

0 = 90Counterclockwise and Vertical Flip (default)
1 = 90Clockwise
2 = 90Counterclockwise
3 = 90Clockwise and Vertical Flip

https://ffmpeg.org/ffmpeg-filters.html#transpose-1

180° (source) :

ffmpeg has no option to do so, but this can be achieved either with 2 consecutive 90° rotations, or with an horizontal flip followed by a vertical flip :

ffmpeg -i input.mp4 -vf "hflip,vflip" -codec:v libx264 -preset slow -crf 20 -codec:a copy flipped.mp4

ffmpeg -i input.mp4 -metadata:s:v rotate="0" -vf "hflip,vflip" -c:v libx264 -crf 23 -acodec copy output.mp4 (source)

Change a video resolution by percentage (source) :

To scale down the video to 25% of its original resolution :

ffmpeg -i inputFile -vf scale=iw*0.25:-1 -codec:a copy outputFile

-codec:a copy allows to stream copy the audio without re-encoding.

Linux workstation : audio and video - How to alter audio and video files on a Linux workstation ?

How to re-tag a .mkv file ?

Situation

Details

Solution