santa hat
deerdeer nosedeer glow
Sign In

Stable Diffusion all samplers compared

Stable Diffusion all samplers compared

In this article I will compare all samplers in Automatic 1111 UI as of June 2023

GPU - RTX 2060 6GB

optimizations: xformers and sdp - scaled dot product

model: Sweet-mix v2.0-flat

image workflow:


Prompt: masterpiece, best quality, 1girl, elf, green eyes, blonde, (standing:1.1), (forest:1.1), pov, perfect eyes, arms at sides, river, portrait, full body, smile, heeled sandals, beautiful landscape, trees

Negative prompt: extra fingers, fewer fingers, bad-hands-5, by bad-artist, (bad eyes:1.2), (misfigured pupils:1.2), (bad clothing:1.3), (undetailed clothing:1.3), ng_deepnegative_v1_75t, verybadimagenegative_v1.3, (nonsensical backrounds:1.2), (bad backrounds:1.2), (bad shadows:1.2), (bad anatomy:1.1)

Steps: 30

Hires fix steps: 15

Seed: 2861404250

Resolution: 1024x1536
CFG scale: 8

Denoising scale: 0.55

All samplers sorted by lowest generation time to highest:


PLMS - 00:56

Euler a - 00:57

DPM++ 2M - 00:57

Euler - 00:58

DPM fast - 00:59

DPM adaptive - 00:59

DPM++ 2M SDE Karras - 01:00

DPM++ 2M SDE - 01:00

DDIM - 01:01

UniPC - 01:05

DPM++ 2M karras - 01:20

DPM2 - 01:26

DPM++ 2S a - 01:28

DPM++ SDE - 01:30

Heun - 01:31

DPM++ SDE Karras - 01:31

LMS Karras - 01:33

DPM++ 2S a karras - 01:33

DPM2 Karras - 01:35

DPM2 a Karras - 02:41

DPM2 a - 02:24

LMS - 08:33 (really odd)


If there is something i got wrong, please let me know. Thanks.

18

Comments