Movie

• Extract keyframes
  • -hide_banner we are using this parameter to hide ffmpeg compilation information
  • -vsync vfr: This is a parameter that tells the filter to use a variable bitrate video synchronization. If we do not use this parameter ffmpeg will fail to find only the keyframes and shoud extract other frames that can be not processed correctly.

ffmpeg -i VIDEO0068.mp4 -vf "select=eq(pict_type\,I)" -vsync vfr thumb%04d.bmp -hide_banner

• Extract one frame per second

ffmpeg -i video.webm -vf fps=1 thumb%04d.jpg -hide_banner

• Extract one frame per 10 seconds

ffmpeg -i video.webm -vf fps=1/10 thumb%04d.jpg -hide_banner

• Extract with different quality / to bmp

ffmpeg -i video.webm -vf fps=1 -quality 90 thumb%04d.jpg -hide_banner

ffmpeg -i video.webm -vf fps=1 thumb%04d.bmp -hide_banner

• Extract only one frame

ffmpeg -i video.webm -ss 00:00:07.000 -vframes 1 thumb.jpg

https://en.wikipedia.org/wiki/Chroma_subsampling

• j:a:b - j - horizontal sampling reference (usually 4). a number of chroma samples (Cr, Cb) in the first row of j pixels. b number of changes of chroma samples between firs and second row of j pixels.
• 4:4:4 - (H265, H264) - no chroma subsampling - for every 4 pixels of luma, there are 4 pixels of colour
• 4:2:2 - 50Mbit - (H264) for every 4 pixels of luma, there are 2 pixels of colour. Keying can produce bad edges.
• 4:2:0 - 35MBit - (DVD, DV, JPEG, MPEG1, Blueray) - only alternatign lines are sampled for chroma. Chroma vertical resolution is halved.

ffmpeg -threads 8 -i $1 -vf deshake -c:a copy -c:v utvideo yuv420p $1-deshake.mkv

NOTE: No multithreading.

FFMpeg already contains plugins vidstabdetect and vidstabtransform http://public.hronopik.de/vid.stab/download.php. But latest version (supports multithreading) can be downloaded as statically linked binary from: https://www.johnvansickle.com/ffmpeg/

ffmpeg -i $1 -vf vidstabdetect=shakiness=10:accuracy=15:result="tracefile.trf":show=1 -c:a copy -c:v utvideo ${PIX_FMT} ${OFILE}-vidstab1.mkv
ffmpeg -i $1 -vf vidstabtransform=input="tracefile.trf" -c:a copy -c:v utvideo ${PIX_FMT} ${OFILE}-vidstab2.mkv

NOTE: output from pass1 is also written as video to see motion vectors.

Use UT Video codec as better alternative to HUFYUV: https://en.wikipedia.org/wiki/Ut_Video_Codec_Suite

• Download for Windows: https://www.videohelp.com/software/Ut-Video-Codec-Suite
• Supported by ffmpeg

ffmpeg -i $1 -vf deshake -c:a copy -c:v utvideo -pix_fmt yuv422p $1-deshake.mkv

Comparison of file size:

The same video encoded using utvideo @422p is 78% of huffyuv.

Highest quality (qscale 4-5 is far enough)

-c:v mpeg2video -qscale:v 2

DVD's VOBS can be simply concatenated to one big file:

cat ./VIDEO_TS/*.VOB | ffmpeg -i - <out_name>.<out_format>

#!/bin/bash -eu
 
VOBS=""
while read VOB; do
    echo ${VOB}
    if [ -n "${VOBS}" ]; then
	VOBS="${VOBS}|"
    fi
    VOBS="${VOBS}${VOB}"
done < <(ls -1 *.VOB)
 
ffmpeg -i "concat:${VOBS}" -map 0:0 -map 0:1 -map 0:2 -c copy out.ts

More: https://newspaint.wordpress.com/2016/07/27/ripping-a-video-from-dvd-using-ffmpeg/

CRF (Constant Rate Factor), 0..51 (0-lossless)

• 17..23 very good, no blocking efect during fast movement
• 23 - default

-c:v libx264 -crf 25

-c:v libx264 -b:v 1024k

2.5Mbps bitrate with tolerance 300k

-b 2500k -bt 300k

to half of size (using video filter https://ffmpeg.org/ffmpeg-filters.html#scale

-vf scale=w=iw/2:h=ih/2

1 channel audio (mono), audio quality 7 https://trac.ffmpeg.org/wiki/Encode/MP3

-ac 1 -c:a libmp3lame -q:a 7

ffmpeg -i input.ts -vcodec copy -sameq -acodec copy -f matroska output.ts

Create mono mp3 audio stream:

ffmpeg -i input_file.avi -c:v copy -ac 1 -c:a libmp3lame -q:a 7 output.avi

ffmpeg -i example.mkv -c copy -an example-nosound.mkv

ffmpeg -i input.mp4 -b 1000000 output.mp4
ffmpeg -i input.mp4 -vcodec libx264 -crf 20 output.mp4

ffmpeg -i input.mp4 -c:v libx264 -crf 23 output.mp4
ffmpeg -i input.ts -c:a copy -c:v libx264 -crf 18 -preset veryfast NL.mp4
ffmpeg -i input.mkv -c:v copy -c:a 

Limit CPU usage to 150% (2 CPUs)

cpulimit -z -e ffmpeg -l 150

Cut input from 11:20 to 1:45:50:

Example below (option order -i and -ss) will decode input from beginning frame by frame and starts processing at 11:20 and stops at 1:45:50.

ffmpeg -i input.mp4 -ss 11:20 -to 1:45:50 -c:a copy -c:v copy  output.mp4

Next example (option order -ss and then -i) will seek to position 11:20 using keyframes and start processing for 1:34:30.

ffmpeg -ss 11:20 -i input.mp4 -to 1:34:30 -c:a copy -c:v copy  output.mp4

This is faster solution.

In both examples video codes is “copy” so seeking works using I-frames, so there is no speed difference. Additionally with “copy”, it is possibilty to select time range without keyframes, so only audio will be copied and video will start from next valid key frame.

Copy chosen streams:

ffmpeg -i test.ts -to 5:47 -map 0:0 -map0:1 -map 0:4 -c copy cut.mkv

Copy chosen streams and recompress video stream:

ffmpeg -i test.ts -to 5:47 -map 0:0 -map0:1 -map 0:4 -c copy -c:v libx264 -crf 25 cut.mkv

Copy chosen streams and resize to 50%, recompress video stream:

ffmpeg -i test.ts -to 5:47 -map 0:0 -map0:1 -map 0:4 -c copy -c:v libx264 -crf 25 -vf scale=w=iw/2:h=ih/2 cut.mkv

 ffmpeg -i test.ts -vf yadif -vcodec ... -acodec ... test.mp4

vidia

https://www.maketecheasier.com/convert-video-to-mp4-handbrake-linux/

* deprecated pixel format used, make sure you did set range correctly

• when using yuv420p pixel format