Lately I've been feeling a little nostalgic towards the style I had in some of my older gens (namely ones from around the time of post #129513, post #129519 and post #129220). At the same time, despite really having looked forward to the LoRA I use in my current style, I've started to get a bit fatigued of some of its' quirks. With that said though, there are still some thing about it I like, so I thought of going back to my old style, but throwing the LoRA in there at a reduced intensity to see how it'd turn out.
In truth it's not actually that different from my current style. My older style used 3 artist tags (technically 4, but turns out one of them was a red herring that didn't work) with 2 LoRAs and the shiny_skin tag, because I liked it at the time. My current one is those same 3 artist tags, but with those 2 LoRAs and the shiny_skin tag replaced by that single LoRA I mentioned, at 0.8 intensity. Both of them also have different lineart-related tags, but that's irrelevant since the result is the same, it's just that I found a more consistent way of doing it now.
For the test itself, I've decided to gen sets of 5 images with the same T2I seed, where the only differences between them is the intensity of the LoRA. I couldn't decide between 0.2 and 0.4 intensity, but upon recieving some feedback, I decided to go with 0.4 intensity. I also ended up removing the shiny_skin tag because it was a bit much when combined with the LoRA.
I picked Sagume as one of my subjects, because she's always been my go-to for style tests.
i'm envious of anyone who can settle on an artstyle. i'm constantly trying very different ones, throwing all kinds of styles into a lora and seeing how it turns out, etc. yours looks great!
Thanks! Experimenting with artstyles can be really fun, some people just have so much fun with it that they'd rather keep doing it instead of settling, which is understandable. Settling on the other hand is probably only something you'd want to do if you want to regularly upload and have a bit of "you" in your gens, something akin to a signature I guess, even if at a surface level it is a bit hypocritical to describe it like that considering it's AI we're talking about.
I think a surefire way to settle on a style would be to throw in a handful (or less) of artists you like into the prompt box and seeing what you get once you play around with the tag intensities for each of them.
March 2026 Style Tests (Sagume)
Lately I've been feeling a little nostalgic towards the style I had in some of my older gens (namely ones from around the time of post #129513, post #129519 and post #129220). At the same time, despite really having looked forward to the LoRA I use in my current style, I've started to get a bit fatigued of some of its' quirks. With that said though, there are still some thing about it I like, so I thought of going back to my old style, but throwing the LoRA in there at a reduced intensity to see how it'd turn out.
In truth it's not actually that different from my current style. My older style used 3 artist tags (technically 4, but turns out one of them was a red herring that didn't work) with 2 LoRAs and the shiny_skin tag, because I liked it at the time. My current one is those same 3 artist tags, but with those 2 LoRAs and the shiny_skin tag replaced by that single LoRA I mentioned, at 0.8 intensity. Both of them also have different lineart-related tags, but that's irrelevant since the result is the same, it's just that I found a more consistent way of doing it now.
For the test itself, I've decided to gen sets of 5 images with the same T2I seed, where the only differences between them is the intensity of the LoRA. I couldn't decide between 0.2 and 0.4 intensity, but upon recieving some feedback, I decided to go with 0.4 intensity. I also ended up removing the shiny_skin tag because it was a bit much when combined with the LoRA.
I picked Sagume as one of my subjects, because she's always been my go-to for style tests.