My eleventh successful embedding train. All thanks to @JernauGurgeh for their awesome guide and the help in fine tuning what I was doing wrong.

Edit - Based on some feedback, I'm recommending that the TI weight be 0.9-0.99 at most depending on the checkpoint. I don't know what is going on with this one and I will revisit her over the weekend. I have some additional training images I can try another run on. I used a few more epochs and training images than I did on my others and I'm curious if that caused some of these issues.

Brandi Braids is an American pornographic performer of mixed racial descent active from 2020-present (born 1997). Training data had makeup and jewelry and seems to work with the prompts.

Nudity is possible due to the subject matter. Please prompt accordingly.

This version is the 135-step of the training run. I do have another version, trained slightly differently that I may upload at a later date.

SPECIAL NOTE: I had some of my testing images generate with very Asian features for her. On further testing I learned that this TI is very, very sensitive to prompt weights, even weights outside of the TI itself. i.e. I had several tokens weighted at 1.2 and 1.3 and had to drop these back to 1.1-ish to get consistent pictures. It may be a function of my crappy graphics card, or it may be a feature of the TI. I found that adding "asian" to the negative prompt helped but feel that has some ethical issues given her ancestry.

All testing was done on Jernaumix V4 https://civarchive.com/models/163310/jernaumix?modelVersionId=327618.

Please do not post content that violates CivitAI's policies. Using this resource to generate adult content may violate the law in your area and I take no responsibility or liability for what you use this for after downloading.