Skip to content

Latest commit

 

History

History
91 lines (54 loc) · 3.02 KB

README.md

File metadata and controls

91 lines (54 loc) · 3.02 KB

stlizer

stlizer is a fast conservative video stabilizer.

It basically just translates or rotates the video frames. It means that lens distortion or motion blur cannot be fixed, but there is no need to worry about unwanted distortions or artifacts caused by algorithms or AI models.

I use this software just for watching videos or for pre-processing for other software, so this is the best specification.

example1.mp4

Realworld video example from DeepStab dataset.

Usage

python3 -m stlizer -i input_video.mp4 -o output_video.mp4

(If you are using nunif-windows-package on Windows, run nunif-prompt.bat and enter the command from the console.)

Show all options

python3 -m stlizer -h

Only important options are described below.

--border option

You can specify out-of-frame processing with --border option.

--border black

Simply fill with black color. The output resolution will be the same as the input video.

black.mp4

--border crop

Crop with the fixed ratio specified by --padding option (default 0.05 = -5% border cut).

The output resolution will be smaller then the input video.

crop.mp4

--border expand

Simply fill with black color. Add border padding specified by --padding(default 0.05 = 5% addtional border).

The output resolution will be larger then the input video. There is less potential for some pixels of the frame to be cut off.

expand.mp4

--border outpaint

Fill with coarse blurry outpainting.

outpaint.mp4

This would be suitable for input to other software such as iw3. However, I have not tested it yet.

--border expand_outpaint

outpaint version of expand.

expand_outpaint.mp4

Smoothing Strength

You can specify the smoothing strength with --smoothing option (2 by default).

If a higher value is specified, the original camera trajectory will not be respected and the out-of-frame area will increase.

About Video Analysis Cache (*.stlizer)

*.stilizer file is a cache file for pre-computed camera trajectory. It does not include raw file names or any privacy-sensitive data.

If this file exists, pass1 and pass2 can be skipped, when the options are compatible.

The storage location depends on OS.

Linux: ~/.cache/stlizer
macOS: ~/Library/Caches/stlizer
Windows: %APPDATA%\\Local\\nunif\\stlizer\\Cache

When --disable-cache option is specified, the cache file is ignored and not created.

All cache files can be deleted with the following command.

python3 -m stlizer.purge_cache