Tessa Fowler
Tessa Fowler v1
First and final version. 53 images of no glasses, 43 images of glasses, ranging from tight chin-forehead crop, shoulders, to full body. Trained in WebUI 2.1.6 using Dreambooth extension on SDv1.5.
Training details:
Steps per image: 100 (with several restarts)
Batch Size: 3, GA: 3
LR Sched: constant
Unet LR: 0.00002
Text LR: 0.000005
Regularization using 'sexywoman' images from HF.
Max resolution 1024 using buckets.
Concepts using 'photo of <trigger word> woman'
Dim, Alpha 64
Tessa Fowler, pinup glamour model LoRa.
Use trigger words 'Tessa Fowler', and optionally 'wearing glasses' to add her signature black frames. Try [red hair] or [red lipstick] to accent. Works well 0.7-.85 weight. LoRa is nsfw so add to negative prompt.
See version for training details.