Trained on Tsubaki, 13 images.
DiT V2: replace "eye necklace" with "eye-shaped necklace", just to make her necklace look more accurate.
Applies for DiT V1: Important tip: If you use prompt helper ( automatically turned on for Tsubaki, you can temporarily turn it off manually in the settings), I recommend that you write your prompts, or copy the prompt AI writes and modify it to your liking. Because, for some reason, the prompt helper thinks Aki is a dark skinned female. No idea why, I did not use any image that would feature any other character, besides Aki, or modify her skin. Hope you enjoy, and sorry for the inconvenience. Just manually modify your prompt, and it will be alright. That's what I did, and it was fine afterwards.
I use Euler A. 28 steps. Because the results are pretty good, I think you can generate 3 images in parallel, and you will still be fine. Unlike SDXL, DiT is pretty consistent from my initial experience. Of course, if you demand a lot of detail, then maybe go cheap first.
I use 3, 4.5 and 6CFG. 0.7 Rescale CFG, same as every DiT V1 LoRAs I use.
Thumbnail model: Tsubaki
