>>18506547In auto1111 there should be a tab for "training" or "checkpoint training". you'll need a base model (usually SD 1.5 is popular), some images (cropped to the correct dimensions), and the images need to be tagged. Then just adjust some settings ("hyperparameters") and let the GPU do the rest of the work.
Model merging is much simpler. It's like taking models you like and mixing them like mixing paints. But beware of getting artifacts compounded from this.
