Generative AI for Art
by Ashok Balasubramanian Updated: Sep 6, 2022
Guide Kit
AI-generated artwork wins the top prize in a U.S. art competition! Jason Allen's "Théâtre D'opéra Spatial" or 'Space Opera Theater' won the top prize at the Colorado State Fair's fine art competition in the "digital arts/digitally-manipulated photography" category. While the category allowed digital art, this issue has ignited fierce debate on A.I. generated content. We have all been accustomed to chatbots talking to us in natural language or text editors used in blogs. Creativity is the hallmark of human evolution! The ability to create art is one of the defining characteristics of evolution. The past generation of automation technologies went after repetitive manual tasks and wasn't seen as much of a threat. Generative A.I. technologies promise higher-level cognitive task capabilities such as writing, coding, video, and art. So naturally, this will usher in industrial revolution scale debates on the balance of using A.I. vs human economic value add. The other dimension that is also playing out is copyright. There are at least three parties involved in making A.I. art. The millions of images and their authors, the model's technology provider, and the user who generated the art. In August, a U.S. appeals court affirmed that an artificial intelligence system could not be an inventor under United States patent law, noting that the inventor must be a natural person. Authors of licenses such as CreativeML Open RAIL-M claim no rights on user-generated outputs. Though the product created by the engine is not patentable, it is unique. How unique can derivative work be, and can it be considered innovative? That is the maturity curve that generative A.I. has to scale. After all, we humans learn from instruction, infer from different sources and then reflect those in our innovation! While this journey evolves, here are interesting open source libraries that will help you generate art using A.I.
imagen-pytorchby lucidrains
Implementation of Imagen, Google's Text-to-Image Neural Network, in Pytorch
imagen-pytorchby lucidrains
Python
6930
Version:1.25.4
License: Permissive (MIT)
deep-dazeby lucidrains
Simple command line tool for text to image generation using OpenAI's CLIP and Siren (Implicit neural representation network). Technique was originally created by https://twitter.com/advadnoun
deep-dazeby lucidrains
Python
4385
Version:0.11.1
License: Permissive (MIT)
stable-diffusionby CompVis
A latent text-to-image diffusion model
stable-diffusionby CompVis
Jupyter Notebook
55328
Version:Current
License: Others (Non-SPDX)
glide-text2imby openai
GLIDE: a diffusion-based text-conditional image synthesis model
glide-text2imby openai
Python
3195
Version:Current
License: Permissive (MIT)
DALLE2-pytorchby lucidrains
Implementation of DALL-E 2, OpenAI's updated text-to-image synthesis neural network, in Pytorch
DALLE2-pytorchby lucidrains
Python
9765
Version:1.14.2
License: Permissive (MIT)
dalle-miniby borisdayma
DALL·E Mini - Generate images from a text prompt
dalle-miniby borisdayma
Python
14133
Version:v0.1.1
License: Permissive (Apache-2.0)
aphantasiaby eps696
CLIP + FFT/DWT/RGB = text to image/video
aphantasiaby eps696
Python
721
Version:v2.5
License: Permissive (MIT)
disco-diffusionby alembics
disco-diffusionby alembics
Jupyter Notebook
7233
Version:v5.4.0
License: Others (Non-SPDX)