Set of tools to facilitate prompt design for AI art
- image manager that associates prompt components with images
- simple data mining tools (e.g. n-grams) for extracting commonly used prompt components
- NLPaugs for constructing variations on prompts
- image captioning tools for identifying prompts associated with one or more images
- mechanisms for organizing, categorizing and composing prompt components and templates
- database of useful prompt chunks (e.g. artist names, materials, templates, vitamins for inducing realism or styles, etc.)
- EDA visualizations for examining embedding space topology, clustering, and distributional features of prompts (inliers, outliers, aethetic scores, etc)
- Anticipate that primary users will be non-coder artists and colab users
- Will need to ship with some sort of simple front-end (i.e. a notebook)
- Design for use with notebooks, perhaps include a notebook of pre-designed colab forms or compatible widgets
Some tinkering with the simulacrabot dataset: https://colab.research.google.com/drive/1DDGxDIEuxFCJskP2Mo1DN0wNHqS-QCD7#scrollTo=gOO8cXtr47MY