v1.1
About v1.1: Not trained by me, but by a fellow traveler NerissG with his experimental training setup. Turned out absolutely great! This is the winner. Go, give him a hug!
Trigger: @vinne or Art by @vinne
Strength: 0.8-1.0
Sampler: dpmpp_sde / dpmpp_sde_gpu (might be faster and more optimized on your hardware)
Scheduler: Simple
Steps: 16 (16 steps with dpmpp_sde is like 32 euler steps)
CFG: 4
or
with dmd2 turbo lora 8 steps CFG 1 with euler simple
or
dmd2 lora at 0.5 weight ,16 euler steps CFG 2-3
You could start a prompt by: Art by @vinne with visible paper texture and dark muted colors, painterly film grain.
v1.0
After testing a bit around, here is my opinion: Download v1.0 and also ep14 from the additional components. v1.0 is a very nice allrounder - not entirely flat and not too 2.5dish. Ep14 is more on point with shorter prompts, a bit more 2.5dish, so more depths.
EP18 is probably a bit overfit and EP17 is the flattest one.
Trigger: @vinne
Strength: 0.8
Turbo LoRa Strength: 0.5
Optional prompt: The artistic style is painterly with a heavy, gritty texture throughout, characterized by a muted color palette.
Steps: 15
CFG 3
Euler+Simple
And use the Layerskipnode from the official workflow
EP14 & Ep18 and Ep17 are also interesting versions, added them to the components list on the right side. Overall I think v1.0 won the most comparisons on same seed generations. 14 and 18 are designed with a "v2" at the beginning of the filename and they are rank32 trained, so from a different training. Ep 17 is like v1.0 rank 16 and trained in the same training.