0.9-1.1 work best
version log:
v1: downloaded all existing artwork online and train to get basic raw version
v2: using v1 lora generated image to further more enhance database of lora training
v3: using v1 or v2 or (v1 and v2) lora generated image to enhance database of lora training
v4: repeat all above and choose only high quality image in lora training from v1,v2,v3 and all combination of old lora version use. currently stable(kinda)
