A LORA IS COMING

The Kohya_ss repo is currently having some issues with LoRA extraction from V1.5 DreamBooth checkpoints. The issue is currently fixed for SDXL, but not V1.5. While I'm waiting for a fix, and to get into SDXL LoRAs (coming soon), here is the larger V1.5 checkpoint file.

Eventually I will be producing exclusively SDXL LoRAs, but I have a long way to go with that yet.

Shivam Dreambooth checkpoint of British actress Stephanie Beacham, a popular face on TV and in movies in the UK in the 1970s and 1980s - when she also became a major US soap star.

Uses 50 ref images and 1989 (real) reg images, trained for 8,000 steps at a batch size of 1, for 9 epochs.

I generally use these models in complex workflows, inpainting faces after initial T2Is, and using ControlNet extensively. So if you're hoping for one-click prompt magic, my models aren't data-curated with this in mind, but rather as tools for traditional workflows that use Photoshop and other older methods.

This model may well produce subject-accurate NSFW output unless countered in negative prompts.