vid.stab is a video stabliser for
Run a simple command in a shell, end up with stabilised video.
Commands are easily scripted; run batches overnight when nothing else is happening, or when you go to work. Run it on a server. Run it on a cluster.
This is a fully automated stabilisation process. You don't get to open a window and see how the stabilisation will affect the video, or do manual overrides in the middle of a clip.
The Slackermedia build of ffmpeg has
vidstab enabled. Other builds of
ffmpeg may not have this feature enabled, so if you used another
ffmpeg build or built it yourself, you should double-check that it is active:
$ ffmpeg --version
Look in the output for the
–enable-libvidstab flag. If it is set, then your installed version of
ffmpeg can use
Vid.stab is a two-step process; first video must be analysed, and then it can be stabilised.
Fine adjustments probably need to be made depending on the severity of the shaking in the video. Read more about the different options on https://github.com/georgmartius/vid.stab, and as always, practise makes perfect. Learn to identify the different “kinds” of shakes by stabilising as many clips as you can; take notes on the results, and try different techniques.
First, analyse a video file, using the filter
$ ffmpeg -i foo.mp4 -vf vidstabdetect=shakiness=10:accuracy=15:result="transforms.trf"
There is no video output from this command, but data about the video is written to the file
Use this data to stabilise the video, applying a slight zoom so that the transforms around the edges of the screen are not visible:
$ ffmpeg -i foo.mp4 -vf vidstabtransform=zoom=5:input="mytransforms.trf" fooStabilized.mp4