<p>I tried to train a maxwell parish style but didn't give it enough steps of training. I calculated 125 epochs but I think the real challenge was that it was difficult to scrape together more than like 15 images of good resolution. So the result is still a style Lora that effect lighting and is still neat to use when cranked up in power. Certain language triggers it better than other language , mentioning yellow colors and red colors and blue skies, tree's trigger also. The trigger word itself which is "parrish light style" doesn't seem work well unless you start the prompt with , "Parrish Light style painting, then mention red rocks, trees, yellow tones, blue cloudy skies, or mention all together then it will pop something into place thats very pretty. I thought I'd share it since I do use with my other models and such. This way I can give the lora some credit.  It does make nice image,  all image are made using my Zimage bean model with  this lora at 1.0 power and refined in turbo for last 4 steps after an initial 16 steps of base generation</p>

Maxwell Parrish Finicky

<p>This will work on zimage base or turbo.   Because turbo isn't as flexible it works best on base.    If you use a two pass workflow like me,  then load it into base and turbo in your workflow.  This go's for any of my lora's   They are completely interchangeable between base and turbo zimage models.   </p>

Files with SHA256: 873fc3c282f128e2d62db653321d9d7f42789498a1e1988267c4018a86dd5249