Training Textual Inversion for Stable Diffusion | Customizable AI Image Generation

Описание: A quick look at training Textual Inversion for Stable Diffusion.

This is a method of training a phrase to be associated with a set of images, which can then be applied to the stable diffusion model to generate similar output.

It's not quite retraining the model, but the results are similar, and require less input and training time. The training feature was recently added (10/2022) to Automatic1111's Stable Diffusion WebUI, and is pretty easy to use. Here I attempt to train it to generate artwork similar to concept images from the cancelled Fallout Van Buren project.