THIS IS THE NEW (DECEMBER 2023) V2 IMPROVED LORA FOR ADRIENNE BARBEAU

An entirely new dataset was created, and the new LoRA was trained from scratch in Kohya, three times, and the best results merged at 100% each, according to this method, which I also used with great success for Slave Leia, Jamie Lee Curtis, and Beverly D'Angelo.

The new Barbeau LoRA is far more disentangled and flexible, and a notable improvement on the prior version (V1). Please be sure to choose the V2 version before downloading.

If you append the word 'young' to the trigger, Barbeau's default curly perm (1979-87) may straighten out a bit and go younger. The range of data used for this covers quite a period of time.

Due to Civit's extreme policies on photos of celeb AI models, I am only supplying one face photo for this model, but it is capable of much more.

REMEMBER THAT THIS MODEL REQUIRES A LORA STRENGTH OF NO MORE THAN 0.1!

-- VB, Monday, December 4, 2023

This model is very likely to produce NSFW and nude renderings unless counter-prompted.

OLDER (OUTDATED) TEXT

THIS IS THE LORA VERSION OF THE ORIGINAL 2GB CHECKPOINT FILE).

CLICK HERE FOR A WIDER RANGE OF PREVIEW IMAGES

Shivam Dreambooth>LORA of American actress Adrienne Barbeau, using 84 ref images and 1980 (real) reg images, for 12,000 steps.

The period covered is around 1975-1982. To get a generation of the younger version, put the word young in front of the key prompt, and use brackets to control the degree of youth.

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.