INFORMATION ABOUT V2

NEW VERSION - Helen Mirren V2 LoRA. (March 2024)

The new version is trained on 75 images using the multitrain/merge method. It is far more flexible than the old version.

This one was trained from scratch in Kohya.

It has a new trigger: mirrenex

It requires a LoRA strength of around 0.14 (not 0.1 - the civit.ai system will not allow fractions when captioning the model, so that is an error. 0.1 is too low)

It is quite likely (as was the previous version) to produce NSFW results, unless counter-prompted.

OLD INFORMATION (PREVIOUS VERSION):

Shivam Dreambooth>LORA of British actress and 1970s/80s sex symbol Helen Mirren, using 316 ref images and 1400 (real) reg images. This model is a merge between two of my own models of Mirren, from the period 1977-84 approximately. The first model is body-centered, with high detail on face, and the second concentrates on producing camera-facing poses. The average number of steps between the models was 12,000.

The period covered is around 1978-1985. 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.

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