The ComfyUI workflow is attached!
bigASP is a photorealistic SDXL finetune on over 6 million photos. The creator has documented usage extensively on the civitai page so it’s worth reading! I think the primary strength of the model for me is that it can generate a wider variety of bodies and faces than most other finetunes. The tradeoff is that you have to work harder to get “pretty” aesthetically-pleasing images. I think it’s more than worth it though since you can get outputs with this model that you simply can’t get with others. If you’re interested, the full tag list is here and the natural language n-gram list is here.
I've uploaded my testing grids to Drive here (huge images). I've summarized my findings below, but the grids should theoretically validate my recommendations if you're interested.
Sampler: DPM++ 3M SDE
Scheduler: Exponential
Steps: 40
CFG: 3.0
Recommended prompt for pro photo shoot look
Positive: score_7_up, photo \(medium\), [main prompt], realistic
Negative: score_1_up, score_2_up, score_3_up, score_4_up, jpeg artifacts
Recommended prompt for amateur look
Positive: score_2_up_, photo \(medium\), [main prompt], nsfw, onlyfans, reddit, realistic
Negative: [leave blank]
Random observations
Photo \(medium\) seems to slightly improve anatomy
Watermark and signature seem to improve anatomy while also making the photo look less amateur
Nsfw, onlyfans, and reddit (as well as use of subreddits) seem to improve anatomy while still looking amateur.
Realistic seems to improve anatomy