Any advice for training single or few images LoRa training?

Someone has some tips for training a LoRa for character that have a single image or few, i tried spliting the image and adding flip versions , i got some ok-ish result , but i will like to know what methods some may use.

Try to flip, zoom in, zoom out, adding some noise, rotate, stretch,etc sometimes the results are pretty good but with some settings that fits your dataset