Noob here, I spent a little time picking up some pointers from the thread and wiki, thanks everyone for all the info and inspiration. After messing around for awhile I finally started getting decent results with sporty Pekora. I'm using hll-5c with hll-5c-lora, is there a more reliable model? This is definitely better than any other AI thing I've ever done, but it seems a bit messy compared to the results itt. Some takeaways I've found: all the sliders make a huge difference, sample method makes a huge difference, subject matter makes a difference, hands are menace. Would love tips on technique.
I can't get my seed and prompt to save properly in the filename, so here is what I used:
>>823590005 best quality, masterpiece, ultra-detailed, illustration, perfect anatomy, expressive face, v-shaped eyebrows, volumetric shadows, ( ( usada pekora, hololive) ), rabbit ears sticking up, cute, smug, smiling, full body wearing sports bra and leggings showing off nice ass <lora:hll-5c-lora:1>Sampling method: DPM adaptive, Sampling steps: 75
Upscaler: R-ESRGAN 4x+ Anime6B, Hires steps: 15, Denoising strength: 0.45, Upscale by 2
CFG Scale: 9
Refiner: anything-v3-full, Switch at: 0.8
