Settings from my Images : ~ Euler A, 8 Steps ~ Hires fix : 25 steps, 0.55 Strength ~ Config : 2 , same as hires CFG as well 2. ~ Probably will collide with other LoRAs but might worth the test.
Illustrious, Different style, LoRA-merged in, it works with or without negative prompts ( maybe ). Settings from my Images : ~ Euler A, 8 Steps ~ Hires fix : 25 steps, 0.55 Strength ~ Config : 2 , same as hires CFG as well 2. ~ Probably will collide with other LoRAs but might worth the test.
Im a big fan of VPX Hyper and Turbo models... so i had to try this one. The things this merge can achieve with just a few steps are great. Not always but even with 8-12 steps you get good hands. Amazing work.
Works well with few steps. At 8 steps all the complex interactions in my prompt like where the cup should be and where the laptop should face was hard to get right so I increased it to 16.
Negative prompting "overexposure" can help with some of the side effects of extra steps.
Great work! But even if (((uncensored))) is included in the negative prompt, the generated images still have many black lines, which is quite annoying.
nice work! The performance of this model is absolutely stunning. I'm curious to know which LoRA has been merged and if there are any plans to release a vpred version?