Squeebo's Layered Art Randomizer

Prerequisites

• Install python 3.x
• Use pip to install pillow:
  pip install pillow

Setup

The primary requirement for using these scripts is to unpack and organize files into some folder. Because I'm anal, that folder is usually 1-layers. In this folder, you should organize things into unique areas of the PFP. For example:

1-Backgrounds
2-Skin (or 2-Body)
3-Clothes
4-Mask
5-Hair
6-Hat
7-Accessory

This is optional, but will help 1-loader.py group things more intelligently. Manually editing the feature values in step 1 will also create groups.

Sometimes art goes through a few revisions, and I don't want to be the guy that lost the art. Because of this, I usually make another folder called 0-sources which holds "raw" layers. After each round of art is loaded, I use Beyond Compare to identify which layers have changed. With the sources and layers organized, it's easier to adapt to unruly art projects.

1. loader.py

usage: python 1-loader.py path/to/layers

path/to/layers - Indicate where the layers are organized. If you followed the Setup, this value should be 1-layers

This tool will load all of the files it finds into "traits.csv" with the following columns:

• z the height of the layer; lower = further back, higher = toward the front
• feature: the grouping name (key) that will be tracked in the python code
• expression: the grouping value that will be tracked in the python code
• mils: defaults to 1; an integer relative value to control randomization
• link:feature" advanced; use with link:expression to link the current layer to another
• link:expression: advanced; use with link:feature to link the current layer to another
• is_metadata: defaults to 1 (yes); allows you to toggle this layer from creating metadata
• display_type: defaults to "string"; allows this trait to use other display types in the Metadata Standards
• trait_type: the trait type within JSON metadata
• value: the trait value within JSON metadata; title casing will be applied
• path: the relative path to the image/resource
• display_type: defaults to "string"; used for distinct OpenSea attribute types

See samples/traits.csv for an example

After you have applied some organization to traits.csv, move on to step 2 and see what you've created.

2. generator.py

usage: python 2-generator.py

Execute this file next to "traits.csv" to perform several generative actions:

1. Load all of "traits.csv" into memory including extra columns. To use extra columns, they must have a column header.
2. Generate TraitGenerator.QUANTITY unique combinations according to the mils values
3. Generate images and JSONs matching the items created in #2

Configuration

• TraitGenerator.BASE_TRAITS: Indicates which feature (keys) are used for randomization
• TraitGenerator.LOW_MEM: Turned off layer caching. Generator will be slower, but won't crash from running out of memory
• TraitGenerator.METADATA_FORMAT: Provides templated data for the name, description, and image of the metadata
• TraitGenerator.QUANTITY: The quantity of unique combinations to generate
• TraitGenerator.RESIZE: If this is a 2-tuple, resizes the layers to these dimensions to expedite processing

Troubleshooting

• KeyError: 'Overlay Top Most Layer'
  If you receive this error, it means the script cannot find the feature that you specified in BASE_TRAITS. Check the the script and CSV for spaces, tabs, or non-ascii characters.