Initial release of Classically Abby. In Automatic1111's webui, you trigger textual inversions based on the filename. If you leave the filename as-is, place this file in your "embeddings" directory, then trigger the embedding by including "classically-abby-v1.0" in your prompt:

a portrait of classically-abby-v1.0 woman, wearing a t-shirt, outdoors, etc

This embedding is pretty sensitive to the model on which it is used. I thought I had the best luck using Realistic Vision 1.3 and 1.4. I got some good results with Dreamlike Photoreal v2.0, too.

I generated the examples before I renamed the file, so if you copy the prompts, make sure to update references of "abigail-shapiro-steps-4000" to "classically-abby-v1.0". Sorry for the confusion!

Check version information for training data. Keep your eyes out for the next version!