Prompts recommandés

very aesthetic, masterpiece, best quality, 1girl, absurdres, solo

Prompts négatifs recommandés

worst quality, low quality, bad anatomy, distorted perspective, text, deformed, logo, lowres, watermark, shota, loli, multiple views, signature, artist name, shiny skin

Paramètres recommandés

samplers

DPM++ 2M SDE Karras, DPM++ 2M SDE

steps

35

cfg

3

clip skip

2

resolution

1081x1580, 512x512, 998x1459, 921x1612

other models

Avarice_v0.4 (faa5d8790e), Melkor_V4_Illustrious_20 (5c12d7adff), Soft_Modifier_IllustriousV4 (7574f01a7b), ReddTheRats_Styles_-_Scratch-000013 (2e10f9998c), ecajdeck_content_epoch_5 (2db08281f7), realistic_filter_[IL] (52251b0609)

Paramètres haute résolution recommandés

upscaler

4x_foolhardy_Remacri.pth

denoising strength

1

Conseils

The merge algorithm compares the UNET and CLIP tensors of various models and selects the best ones based on subjective weights.

Future versions will explore more granular adjustments at block level or finer granularity.

Points forts de la version

Experimental merge, no real logic behind the merge outside of picking models I liked and some numbers I thought made sense for weighting them. Future versions will have structured and less subjective evaluations to determine more optimal configurations.

An experimental merge of various models I like, using an algorithm that ideally takes the strengths of each. Thanks to @epiTune for the advice and guidance, I still have a lot to learn and a lot to do, but this is a great start.

NOTE: This model is very much WIP, but results are promising by all the people graciously testing.

Models Used

  • IllustriousXL V2.0 Stable - Used as base merge target

    • 1.0 UNET/1.0 CLIP

  • Rouwei v0.7 eps

    • 1.0 UNET/1.2 CLIP

  • IllumiYume v3.1

    • 1.5 UNET/2.0 CLIP

  • Hassaku v1.3 Style A

    • 1.4 UNET/0.5 CLIP

  • ionsyx v3.0

    • 1.4 UNET/0.8 CLIP

  • Wicked Illustrious Mix v1.1

    • 1.2 UNET/0.5 CLIP

  • mdntIllus Syn v1

    • 1.0 UNET/1.3 CLIP

  • Kokio v2.0

    • 1.2 UNET/0.8 CLIP

  • Diving Illustrious Anime v11

    • 0.5 UNET/0.5 CLIP

  • Bismuth Illustrious Mix v2.0

    • 1.2 UNET/0.8 CLIP

  • NoobAI v1.1 eps

    • 0.7 UNET/1.5 CLIP

  • Unreleased Merge

    • 1.0 UNET/2.0 CLIP

  • Plant Milk Hemp II

    • 1.4 UNET/0.8 CLIP

  • Plant Milk Coconut

    • 1.4 UNET/0.8 CLIP

The above may not make the most sense, but using the algorithm, each model's UNET and CLIP were compared and the best was chosen tensor by tensor, using the above weights as my subjective adjustment to influence how likely they would be chosen. I can't tell you exactly what part of what model is used where, my merge script just chose what it considered the best.

For future versions I will be experimenting with different models, along with more granular adjustments (at the block level, or even more granular than that).

License & Use

SDXL - CreativeML Open RAIL++-M
Illustrious - https://freedevproject.org/faipl-1.0-sd/
NoobAI-XL - https://civitai.com/models/license/1140829

Copied the above license list from Plant Milk's page.

Précédent
extremely detailed (no trigger) - sliders.ntcai.xyz - v1.0
Suivant
NeonFantasy style FLUX, PonyXL - FLUX V2

Détails du modèle

Type de modèle

Checkpoint

Modèle de base

Illustrious

Version du modèle

v0.01

Hash du modèle

416daecfbe

Créateur

Discussion

Veuillez vous log in pour laisser un commentaire.