Introduction

This document contains some of the ffmpeg snippets I use most commonly, and typically in the context of converting research-related image data to video. It is not meant to be exhaustive.

Common parameters

Add input: -i <file(s)>, example: -i color-%01d.png
Add input alphabetically: -pattern_type glob -i '*.png'
Set start frame (before -i): -start_number <number>
Limit frames: -vframes <number>
Overwrite without asking: -y (contraty to -n)
Set ouput codec video/audio: -c:v == -codec:v == -vcodec / -a:v == -codec:a == -acodec
Set bitrate video/audio: -b:v <rate> / -b:a <rate>, examples: -b:v 1400K, -b:v 1M
Set offset between keyframes: -g <offset>
Use video/audio filter: -vf == -filter:v / -af == -filter:a
Set framerate (before input / output to pick target): -r <number>
Format (input/output): -f <fmt>, (use -f lavfi to use filter as input)
Constant Rate Factor (h264,h265): -crf (range: 0–51, 0=lossless, 23=default(h264), 28=default(h265), 51=worst)

Common filters

D->RGB: [in]format=pix_fmts=rgb24[out]: convert greyscale to rgb (required for stacking)
Crop: [in]crop=<w>:<h>:<x>:<y>[out]
Scale: [in]scale=<w>:<h>[out]
Stack: [left][right]hstack[out], [top][bottom]vstack[out]
Vary speed: [in]setpts=0.5*PTS[out]: (remember to half the value of 'vframes', if set), (2x speed, here)
Textbox: drawtext=fontfile=<ttf>: text='<text>': fontcolor=<color>: fontsize=<size>: box=1: boxcolor=<color>: boxborderw=<w>: x=<x>: y=<y>
Concat: ffmpeg -i 1.mp4 2.mp4 3.mp4 4.mp4 -filter_complex "concat=n=4:v=1:a=0"
Width and height divisible by 2: pad=ceil(iw/2)*2:ceil(ih/2)*2

Common formats

For web (more compatible): ffmpeg -i <in>.mp4 -c:v libx264 -crf 23 -profile:v baseline -level 3.0 -pix_fmt yuv420p -movflags faststart <out>.mp4
For web (less compatible): ffmpeg -i <in>.mp4 -c:a libvorbis -c:v libvpx-vp9 -b:v 1400K -g 120 out.webm
For PC: ffmpeg -i in.mp4 -c:a libvorbis -c:v libx264 -crf 32 <out>.mp4
Better decoder support with: -profile:v high -pix_fmt yuv420p -movflags faststart

Examples / techniques

Set background color

Use additional input -f lavfi -i color=c=white:s=<w>x<h> and overlay with it [bg][fg]overlay=shortest=1. You might want to keep the scale of a specific input instead of setting <w> and <h> for the background: [bg][fg]scale2ref[bg][fg]

Remove part of video

Use trim filter [0:v]trim=duration=8[a] get video until second 8, [0:v]trim=start=16,setpts=PTS-STARTPTS[b] get video from second 16 to end (otherwise use end=..), finally, concat results "[a][b]concat" See: https://superuser.com/questions/681885/how-can-i-remove-multiple-segments-from-a-video-using-ffmpeg

Fade transition of videos

ffmpeg \
-i a.mp4 \
-i b.mp4 \
-filter_complex " \
  color=white:1280x960,fade=out:st=4:d=1[alpha_a]; \
  [0:v][alpha_a]alphamerge[overlay_a]; \
  [1:v][overlay_a]overlay=0:0" out.mp4

Stacking videos with textboxes

ffmpeg \
-i a%01d.png \
-i b%01d.png \
-filter_complex "hstack=inputs=2, \
  drawtext=fontfile=/usr/share/fonts/truetype/ubuntu-font-family/Ubuntu-R.ttf: text='3D labeled': fontcolor=white: fontsize=32: box=1: [email protected]: boxborderw=5: x=100: y=5, \
  drawtext=fontfile=/usr/share/fonts/truetype/ubuntu-font-family/Ubuntu-R.ttf: text='3D colored': fontcolor=white: fontsize=32: box=1: [email protected]: boxborderw=5: x=1060: y=5" \
-vcodec libx264 out.mp4

Extract all frames of video as jpg

# highest quality:
ffmpeg -i <video>.mp4 -qmin 1 -qmax 1 -q:v 1 %06d.jpg
# q normally ranges from 2-31, and it is more typical to use:
ffmpeg -i <video>.mp4 -q:v 2 %06d.jpg

Pipe to ffplay

ffmpeg -f lavfi -i testsrc -f matroska - | ffplay -

tmatsuzawa/ffmpeg-snippets.md