I'll definitely see about adding her. Coincidentally, I am making a Nanako LoRA, but it's not Akiyama.

@veryfirstquestion800 Worry not, she's in my to-do list once I'm done with five other LoRAs (I have two in testing RN and two already have their datasets set up)

@veryfirstquestion800 Oh, if you have the dataset, I can actually guide you to making the LoRA yourself. There's a guide that links you to a Colab that makes the LoRA for you and I wound up making a handy dandy calculator that lets you figure out how many repeats, steps, and all that other stuff you need to do in order to make your LoRA work.

In the Dataset you provided, you have 55 images. Of the Guide's four main "healthy" cooking formulas, it's between the 20 images and 100 images. I recently found out that the closer the number is to 100, the more you want the number of Steps closer to 1500, but risking overstepping it.

So, in your case, when you go into the LoRA Trainer, set the number of Repeats to 5 and chances are the LoRA will cook successfully. You could also try cooking at 6 if you want just a few more extra steps to make the LoRA extra crispy, though you might risk overcooking it.

Anything V5 and I do not have a VAE set up.