FFmpeg subtitles commands

extract-audio-from-video.sh

ffmpeg -i input.mp4   \
       -map 0:1       \
       -c:a copy      \
       -y output.m4a

ffmpeg-flac-multiple-streams.sh

for files in *flac ;        \
  do ffmpeg -i "$files"     \
    -map 0:0                \
    -strict experimental    \
    -c:a aac                \
    -b:a 224k               \
    -y "./m4a/$files.m4a" ; \
  done

ffmpeg-sub.md

Overlay

ffmpeg -i input.mp4 \
       -vf "movie=overlay.png [watermark]; [in][watermark] overlay=main_w-overlay_w-10:10 [out]"
       -c:v h264
       -c:a copy
       output-overlay.mp4

Burned subs

ffmpeg -i output-overlay.mp4 \
       -vf "subtitles=input.srt" \
       -c:v h264 \
       -c:a copy \
       output-overlay-subs.mp4

Overlay et burned subs via filter_complex

ffmpeg -i input.mp4 \
	   -i input.png \
	   -filter_complex "[0:0][1:0]overlay=main_w-overlay_w-10:10,subtitles=input.srt" \
	   -c:v h264 \
	   -c:a copy \
	   output.mp4

Non-burned subs

ffmpeg -i input.mp4 \
	   -sub_charenc UTF-8 \
	   -i input.srt \
	   -vcodec copy \
	   -acodec copy \
	   -scodec mov_text \
	   -metadata:s:s:0 language=fre output.mp4

Scaling

ffmpeg -i input.mp4 \
	   -vf scale=1024:-1 \
	   -c:a copy \
	   output.mp4

Scaling

ffmpeg -i input.mp4 \
	   -s 1024/576 \
	   -c:a copy \
	   output.mp4

Most of the time you can't. And not with ffmpeg

Now maybe you could train a model to extract letters from frames, I suppose we now have the tools to apply ML to that (but I don't have the skills to do it).

is it possible to display subtitles vertically down rather than horizontally across with ffmpeg?