SteampunkHotHiRes

Dataset creation: Trained on 128 HiRes (at least 1024x1408) images created on local SD a1111 with SD1.5 based aZovyaRPG_v2 model and a Steampunk clothing LoRA. Further 4x Upscale of the DS. Dataset overview + Samples:

With simple prompts, persons can have the tendency to "long neck", "stretched torso, arms, legs". To put these in negative prompt then is recommended. The model is quite dominant, if the value is too strong, the LoRA shut be weight reduced and/or put it in middle or end of the prompt. With strong value it gives the steampunk style which it was trained on. With lower value it will still enhance and underline an extravagant fashionstyle, even when it is not directly looking like steampunk. That means it is good to use, even when steampunk is not the target. The further US of the DS was a test how much the res. influences the output. Imo the quality is similar, but a diff. result. Alternative to base model.