/glitch_me

Python module to add some distortion/glitch effects to images.

Primary LanguagePythonMIT LicenseMIT

glitch_me

Screencap of a Shinto gate from the anime movie Your Name that has aesthetically pleasing visual distortions

Python module to add some distortion/glitch effects to images.

Inspired by the work of DataErase.

Examples using the included transforms

(Click images for larger view)

Original Glitched GIF'd
tokyo tokyo glitched tokyo glitched gif
cafe cafe glitched cafe glitched gif
gate gate glitched gate glitched gif

Dependencies

Note: This module may work with older versions of these, but compatibility is only guaranteed on these versions or newer.

Install

Download

git clone https://github.com/noelleleigh/glitch_me.git

Recommended Install Method

Add the glitch_me to your Python scripts available on your path and automatically install dependencies by running:

pip install -e ./glitch_me

No Install Method

If you don't want glitch_me added to your scripts, you can run it from within its folder with:

python -m glitch_me

Usage

Command Line Interface (CLI) Arguments

usage: glitch_me [-h] [-q] [--line_count LINE_COUNT] [-f FRAMES] [-d DURATION]
                 [-b]
                 {still,gif} input output_dir

Add some glitch/distortion effects to images.

positional arguments:
  {still,gif}           Make a still glitched image, or a progressive glitch
                        animation.
  input                 Input image path glob pattern
  output_dir            Path to output directory (files will be saved with
                        "_glitch" suffix)

optional arguments:
  -h, --help            show this help message and exit
  -q, --quiet           Include to not print the paths to the output image(s).
  --line_count LINE_COUNT
                        The vertical resolution you want the glitches to
                        operate at
  -f FRAMES, --frames FRAMES
                        The number of frames you want in your GIF (default:
                        20)
  -d DURATION, --duration DURATION
                        The delay between frames in ms (default: 100)
  -b, --bounce          Include if you want the gif to play backward to the
                        beginning before looping. Doubles frame count.

Examples

Add glitch effects to a single image and save it to its existing directory:

glitch_me still /pictures/image.png /pictures

Add glitch effects to multiple images and save them to a different directory:

glitch_me still /pictures/*.png /pictures/glitched

Add glitch effects to a single image, pixelated to a vertical resolution of 200px:

glitch_me still --line_count 200 /pictures/image.png /pictures

Create a glitch GIF from a single image, pixelated to a vertical resolution of 200px, with 10 frames, 50ms frame delay:

glitch_me gif --line_count 200 -f 10 -d 50 /pictures/image.png /pictures

Deeper Customization

The CLI was written as a thin shell and doesn't expose the many ways this library can modify an image. For fine-grained control over the actual transformations applied by the CLI, you must edit the file sample_transform.py.

In sample_transform.py, you'll find two important objects:

  1. STATIC_TRANSFORM: A list of tuples, each containing a function reference and the arguments to pass into that function. This list of functions tells glitch_me how to transform your image.
  2. GIF_TRANSFORM: A function that takes a value called progress that tells the function how far along it is in the animation. The function returns a list in the same format as STATIC_TRANSFORM that is used to transform a specific animation frame.

These two objects define what happens when you use the still or gif options in the CLI, respectively.

The functions referenced in both objects all have one thing in common: They take in a Pillow.Image object (and extra arguments) and return a Pillow.Image object (presumably modified from the input). If you look at the functions that are there already, you'll see that most of them come from the effects.py file, where a lot of useful glitching functions are defined. You can use any function in there with the pattern function_name(im: ImageType, ...) -> ImageType to add a transformation.

Add a New Transformation Tutorial

Say you were looking through effects.py and decided you wanted to add the effect shift_corruption to the GIF transformations.

  1. First, check the function's docstring (the part under the def ... line surrounded by """triple quotes""") to learn what it does and how to use it.
  2. shift_corruption takes three arguments:
    1. im: a Pillow Image (This one is handled by default in the function that uses STATIC_TRANSFORM and GIF_TRANSFORM)
    2. offset_mag: The furthest you want a pixel row to be shifted (in pixels)
    3. coverage: What percentage of rows will be shifted
  3. Now that we know what the function needs, we need to decide what values to give it. The GIF_TRANSFORM function has a useful variable in it called loop_progress, which will let us make an effect that goes from clean, to glitchy, and back to clean over the course of the GIF animation. We'll do some math so that when loop_progress is 0.0, there's no glitching. When it's 1.0, there's maximum glitching. So if you want your maximum offeset_mag to be 10 pixels, then we would put int(10 * loop_progress) into that argument (the int() is to convert the floating point value to an integer, which is what that argument expects). The same principle applies for coverage.
  4. So the code we'll be adding will look like:
    (effects.shift_corruption, {
        'offset_mag': int(10 * loop_progress), 'coverage': 0.5 * loop_progress
    })
    (I've indented the arguments so the line isn't impractically long.)
  5. Now we have to decide where in the multi-step transformation to add this line. Let's add it after the sin_wave_distortion (don't forget to add a comma to the end!), so the final list will look like:
    return [
        (effects.convert, {'mode': 'RGB'}),
        (effects.pixel_sort, {
            'mask_function':
            lambda val, factor=loop_progress, limit=lum_limit:
                255 if val < limit * factor else 0,
            'reverse': True
        }),
        (effects.sin_wave_distortion, {
            'mag': 5, 'freq': 1, 'phase': -2*pi*progress
        }),
        (effects.shift_corruption, {
            'offset_mag': int(10 * loop_progress), 'coverage': 0.5 * loop_progress
        }),
        (effects.add_noise_bands, {
            'count': int(10 * loop_progress), 'thickness': 10
        }),
        (effects.low_res_blocks, {
            'rows': 10, 'cols': 10, 'cells': int(10 * progress), 'factor': 4
        }),
        (ImageOps.posterize, {'bits': int(5 * (1 - loop_progress) + 3)}),
        (effects.split_color_channels, {'offset': int(5 * loop_progress)}),
        (effects.convert, {'mode': 'P', 'palette': Image.ADAPTIVE}),
    ]
  6. Save your modified sample_transform.py try creating a GIF and admire your distorted, glitchy results!