For best results: I've used these samplers with the following CFG:Steps. This is not an exhaustive list. It's just a place to step off from.
DPM++ SDE Karras/Turbo 1-2.5:3-11
DDIM (any scheduler) .6-5:5-30
Euler/Euler A 1-2.5:30-40
VAE is included
Generally speaking, adding steps increases detail but if you want to use any sort of lora or hiresfix, a higher step count may fry your image. Lower cfg tends to follow the prompt better; higher cfg tends to give more contrast.
ADetailer and Highresfix may require slightly different settings than non-turbo models.
Note: If you had a lora stack for 1.0 or 1.2, you're probably going to need to rebuild it for 1.4.
1.4 Is a solid step forward on the road to 2.0. I've merged in some of my favorite loras. It is still responsive and sharper than ever.
1.2 Is more creative and beautiful than 1.0 and tends to handle hands better. The detail may be more subtle or softer but you can always add more with loras, prompts, or steps if you feel you like you're missing something. Cheers~
Version 1.0 is functionally the same as 0.9t, the only notable difference being more life-like and saturated colors. Please enjoy responsibly.
Latest 0.9t comes with baked VAE and is pruned down to around half the original size.
0.9t represents a huge leap forward over the previous iteration. I've merged in some of the realness of Juggernaug and the responsiveness of OpenDalleV1.1 and crammed it back down to Turbo(merge) speed. And well, try it out.