pschaldenbrand / Text2Video

A fast approach for translating a series of text prompts into a video. The 2022 NeurIPS Workshop on Machine Learning for Creativity and Design
GNU General Public License v3.0
32 stars 3 forks source link

another clip-rgb text2video tool #3

Open eps696 opened 1 year ago

eps696 commented 1 year ago

you may be interested in some details of my method https://github.com/eps696/aphantasia, also optimising direct rgb pixels (as well as image fft/dwt parameterization). it's based on the great Lucent library (originated as Lucid, ported to pytorch), and used their tricks of color representation for better optimization (besides adding many more bells and whistles for image control). it was originally presented on CVPR '2021 creativity workshop (see https://computervisionart.com/pieces2021/aphantasia) and later refined at NeurIPS '2021 (see https://neuripscreativityworkshop.github.io/2021/#/gallery, or completed piece https://vimeo.com/647961702, or few other artworks made with that method, available on vimeo)

pschaldenbrand commented 1 year ago

Thanks! Super cool :)