lwc:multimedia_and_graphics:transcoding

Stablization can be done using transcode as outlined here and with more detail here. It's a 2-step process:

create a .trf file with the information needed for transcoding: transcode -J stabilize=shakiness=1 -i test.mp4

shakiness is a number from 1 - 10. 10 means the video is really shaky so apply the filter with full force. 1 means don't do munch. Default is 5 and works well.

process the .trf file with the transform filter: transcode -J transform=smoothing=1 -i test.mp4 -y xvid4 -w 6000 -o test.avi

smoothing sets a low-pass filter for applying the transform. The higher the smoothing value, the more filtering will appear to be happening (it's like raising the shakiness?) Default is 15 and works well.
-w sets the bitrate

Default values work well for most uses, although boosting the framerate to about 6000 improves quality

somehow the audio gets distorted in the conversion. As a workaround, the audio can be added back in later:

There's more detail on xvid4 transcoding with transcode here

it appears transcode gets “stuck” sleeping at the end of a transcode. There's a bash script hack for that problem.

rip audio: ffmpeg -i “$file” -vn -ac 2 -ar 44100 -f wav $fileBase.wav<HTML>
transcode a video file so it is playable through Windows Media Player (Windows) and Quicktime (MacOS): ffmpeg -i INPUT_FILE -b 1500 -ab 384k -cropleft 20 -cropright 20 OUTPUT.mpg
find framerate: ffmpeg -i $file 2>&1 | sed -n “s/.*, $.*$ fp.*/\1/p”
make video from single image (example): ffmpeg -loop 1 -i image.png -c:v libx264 -t 15 -pix_fmt yuv420p -vf scale=320:240 out.mp4 (source)
make video from a series of images (not necessarily numbered from 0000): cat *.jpeg | ffmpeg -f image2pipe -i - output.mkv then ffmpeg -i output.mkv -codec copy output.mp4 from this source

crop: -filter:v “crop=1100:750:0:0” - crops the video to 1100×750 with starting location 0,0
speed: -filter:v “setpts=0.5*PTS” makes video play 2x as fast. So as not to loose frames set -r 30 for 30 fps
to combine filters, use commas. Example: -filter:v “setpts=0.5*PTS, crop=1100:750:0:0”

Don't leave it up to ffmpeg to pick a codec and quality because the quality will be very bad.
-codec:v libtheora seems pretty solid. Set the quality with qscale: -qscale:v 10 is the best quality whereas -qscale:v 2 is the worst.
-codec:a libvorbis teams up with libtheora well. -qscale:a 10 is the best audio quality