>>105573361
the biggest issue right now is automating captioning doesn't pick out styles very well. loras biggest use case is to ape a style or concept the model is under trained in or doesn't know at all