I've got a question for the experts ...
So far I have successfully used zscale for scaling HDR video,
because it is supposed to be better suited for HDR than the
regular scale filter.
However, I'm now in a situation where I need to change the
scaling dynamically, depending on frame number -- that is,
I have to use "n" (frame number) in the formula for width
and/or height, along with the "eval=frame" option.
Unfortunately, zscale does not support that, so I have to
resort to using the regular scale filter.
In the documentation I noticed that it supports options for
setting the color space (in_color_matrix, out_color_matrix),
so I set those to "bt2020" for HDR.
Here's the complete command, broken up in POSIX shell syntax
using variables for better readability:
H='if(between(n\,47105\,47283)\,2160/(2032/1600)\,2160)'
VF="scale=w=-1:h='$H':eval=frame:in_color_matrix=bt2020:out_color_matrix=bt2020"
VF="${VF},pad=3840:2160:-1:-1:eval=frame"
VF="${VF},crop=h=1600"
HDR="hdr10=1:hdr10-opt=1:repeat-headers=1"
HDR="${HDR}:colorprim=bt2020:transfer=smpte2084:colormatrix=bt2020nc"
HDR="${HDR}:master-display=G(8500,39850)B(6550,2300)R(35400,14600)WP(15635,16450)L(10000000,0)"
HDR="${HDR}:max-cll=0,0"
ffmpeg -i input.mkv -vf "$VF" -map 0 -c copy -c:v libx265 \
-x265-params "$HDR" -pix_fmt yuv420p10le -preset slow -crf 20 \
-c:a ac3 -ac:a 6 -b:a 448k -default_mode infer_no_subs \
output.mkv
Technically, that command works and runs without any errors
or even warning messages. The resulting video looks "okay"
on first sight.
However ... when pausing the video and looking closer at the
screen, I believe there are some small artefacts that are not
in the source video. I'm not 100 % sure if my eyes deceive
me or if it's real. And: If it's real, then is it caused by
the scale filter? Or maybe do I just need to lower the CRF
value, or maybe it's something else? And of course, I still
remember the statement that the scale filter is not really
recommended for HDR (zscale should be used instead, but I
can't do that, as explained above).
Can somebody give a recommendation for this situation?
Is it okay to use the scale filter as in the above example?
Can it be improved? Or -- is there a way to use the zscale
filter with a dynamic formula that depends on frame number?
Maybe I've overlooked something here.
Best regards
-- Oliver
PS: Just in case somebody wonders why I'm doing that in the
first place ... My source video is in 16:9 aspect ratio with
2160 pixel rows, but it contains black bands. I want to crop
the video in order to get rid of the black bands.
In the above example, the actual content is only 1600 pixel
rows, but with one exception: There is a short section (just
a few seconds) where the black bands are smaller. Within that
section, the content occupies 2032 pixel rows.
So, before cropping, I have to scale that section (only that
section, not the rest!) down, so nothing gets lost when I crop
to 1600 pixels. Additionally I have to insert a pad filter
that adds black bands at the sides (left and right) within
that section, so the width of the video stays constant.
_______________________________________________
ffmpeg-user mailing list
ffmpeg-user@ffmpeg.org
https://ffmpeg.org/mailman/listinfo/ffmpeg-user
To unsubscribe, visit link above, or email
ffmpeg-user-requ...@ffmpeg.org with subject "unsubscribe".
