Pre Training

I gathered 39 images of Ciri from the Witcher series. I made sure there were equal pictures of Netflix, Games, and Cosplay. Some of the images were full body as I wanted to retain her face even when zoomed out. I used Blip captioning to generate the filewords and edited each individually to reduce potential hallucinations. I excluded any mention of white hair or blue eyes because I wanted to emphasize them in the embedding, and I made sure to mention clothing articles and weapons in the filewords.

Training

I used 0.005:100:0,0.0025:250,0.001:500,0.0005:1000,0.00025 for my learning rate. I am going for 5K training steps total. I am using a batch size of 1 with Gradient Accumulation Steps set to 3. I am running on a RTX 4090 on the cloud. I am using 12.5 out of 24 GB. The estimated time of completion is 2 hours. For the embedding I am using 10 vectors per token. I switched to SD 1.5 EMA Only model for training.