So I tested some of the Models I downloaded recently with the same variables in a as fair as possible manner. I kept the prompts as open as possible while still achieving something comparable out of all of them. You can clearly make out outliars with distinct styles here and there, as well as see some checkpoints having the same roots in their origin.
The Images have been named after the Model's filename - You can check out all the Images in my Google Drive
The tested models were:
absolutereality_v16
abyssorangemix3AOM3_aom3a1b
aiceberg_15
aingdiffusion_v71
analogMadness_v50
animesh_PrunedV10
anyloraCheckpoint_bakedvaeBlessedFp16
AnythingV5Ink
astroChib_v111
aZovyaPhotoreal_v2
aZovyaRPGArtistTools_v3
babes_20
bbmixAnniversary_v20
beautyfoolReality_v10
calicomix_v60
camelliamixNSFW_v11
CheckpointYesmix_v20
chilloutmix_NiPrunedFp32Fix
colorful_v31
colossusProject_v508Creative
copax2DAnime_v10
copaxtimeless_copaxMelodies
CounterfeitV30_v30
darkSushi25D25D_v30
dawnlightLandscapes_dawnlightl
divineelegancemix_V6
dreamshaper_7
eerieorangemix2Hardcore_eerieorangemix2
envymix_v12
epicrealism_pureEvolutionV3
everythingdiffusion_v20
flat2DAnimerge_v30
foxyansfw_v3
galenaREDUX_v20
ghostmix_v20Novae
grilledLamprey_v10o2n3
hellofantasytime_V122
helloflatcolorful2d_V12
humans_v10
icinpICantBelieveIts_afterburn
interiorSmgs_v10
kaywaii_v25
kaywaii_v27
kotosmix_v10
libramix_v20Baked
lyriel_v16
majicmicFantasy_v20
majicmixRealistic_v6
megaModelWith7000X_23AnimeBranch
megaModelWith7000X_24
meinaalter_v3
meinahentai_v4
meinamix_meinaV11
meinaunreal_v41
mistoonAnime_v20
mistoonSapphire_v20
neverendingDreamNED_v122BakedVae
noosphere_v3
paulEberSAnimeMix_v10
perfectWorld_v4Baked
photon_v1
pinenchantedmixBy_pinenchantedplus
progenuberanalogmix_v10
prp_Amateur
pureRealisticPornPRP_v10
realiaPhotorealistic_v10
realisian_v50
realisticVisionV40_v40VAE
revAnimated_v122
sweetness2_v10
synergix_v10
toonSushi_v10
toonyou_beta5Unstable
uberRealisticPornMerge_arpmv13
universestable_V50
universestable_V50Comics
unstableDiffusers_v3UnyieldingGrimoire
v1-5-pruned
xenogasmNDFWSemiReal_v1
xenomix_v10
While I did not get any NSFW results in my generated pictures, I will not link any of these model pages, because some of them are explicitly made for NSFW art.
Testing parameters (Using ComfyUI+WAS Suite) for the Portraits were:
No LoRA/LoCon/LoHa, embeddings or other CLIP/Model altering methods
Base Image Dimensions: 512x768
Positive Prompt: (masterpiece, best quality, detailed:1.2), 1girl, adult female, ginger hair, small braids, blue eyes, freckles, smile, red short sleeved flannel shirt, knee-length denim shorts, sitting in a colorful flower field, looking at the sky, birds in the sky, dreamy mood, magical aura
Negative Prompt: (child, kid, underage, teen, worst quality, extra digits, watermark, signature:1.2)
CLIP -2
Seed: 133713371337
VAE: vaeFtMse840000Ema_v100
Sampling:
Steps: 40
cfg: 8.000
Sampler: dpmpp_2m_sde
Schedulder: Karras
Denoise: 1.000
VAE decode Upscale x4 (ESRGAN_4x)
Resize x0.5 (lanczos, supersampling)
VAE encode
Sampling:
Steps: 10
cfg: 8.000
Sampler: dpmpp_2m_sde
Schedulder: Karras
Denoise: 0.400
Saving Image as PNG
Final Image Dimensions: 1016x1528
Altered parameters for the Landscape Images:
Base Image Dimensions: 912x512 (Almost 16:9)
Positive Prompt: (masterpiece, best quality, detailed:1.2), a small village, an old bridge, forest, roaming monsters, sunset, mysterious, mystic, magical, dangerous
Negative Prompt: (worst quality, watermark, signature:1.2)
Final Image Dimensions: 1816x1016
Cheers