An experiment in batch-processing series of still frames exported from movies, which are then re-assembled as "filtered", lo-fi movies.
This is a bit laborious in terms of the process, relying on exporting hundreds/thousands of still frames, processing from the command-line, and then making new movies. Did I mention it's experimental? ;)
A few examples:
Included:
-
1-bit Atkinson Dither-style rendering, port of FlickrDithr code.
-
Triangular "Pxl-effect"-style rendering based on Rev Dan Catt's Pxl Effect JS implementation.
Dependencies:
- LearnBoost's node-canvas package.
Assumptions:
Both scripts expect source images to be in the pattern of input/movieXXXX.png (eg. 0000 -> 9999) and write to output/movieXXXX.png. QuickTime 7 will export image sequences with this numbering pattern.
Prep:
- Make, or get, a movie.
- Export frames at <= 720p for "new aesthetic", maybe 640x480 for the Atkinson style. I use QuickTime 7 (pro, $29.99) to export an image sequence from the .mov in this case. save as "movie", PNG format, in the output path. Frames take pattern of moviexxxx.png.
- Run script to process images, 250 at a time to get around max file limitations (dependent on your OS.)
Usage:
To process input/movie0001.png ... input/movie0010.png, for example, with either method:
node flickrdithr-node.js 1 10
node dither-newaesthetic-node.js 1 10
General disclaimer:
Since I don't know node.js, I'm sure I'm doing all sorts of terribly-inefficient things. One fun one: Don't use ranges of 250+ files per execution, because the OS will hit a max-open file handle limit. Not sure why this is, despite trying to close / null files after writing.
Following the FlickrDithr code: "Released without license, as is, with no guarantee, support or anything else."