DarkFunIXL

Update: i am re-publishing this previously un-published model as i missed Tuli and wanted to re-do some of my old pictures using the same prompts.

I will not re-publish the Alexx E. edition. Boy, do i miss DarkFun, didn't realise until too late the he left T__T

Historical description:

This is a merge of the checkpoint and LoRAs used by DarkFun for his storytelling.

V1 - Tuli Edition

Part one of the story - Part two of the story - Part three of the story - I am looking forward reading/seeing more!

Usage:

• Sampler: Euler A - Scheduler: Beta (other choices displayed here)

• CFG: 5.5

• Steps: 36

• Clip skip: 1

• Prompt quality tags: "HDR,best quality,detailed,realistic"

• Negative prompt: "bad quality,low detail,child,kid,lolita,worst quality, bad anatomy, bad hands,anime,lowres,plastic skin"

Update: I have created and uploaded a dataset of 1300 "1girl" in a cyberpunk setting using this checkpoint. It is available here

My next step will be probably first to create an embedding for Tuli, the OC presented in the story, and maybe a LoRA, trained specifically on this model :D

NB: Beware of ADetailer with this model/settings. A complex prompt can't be used as-is for face inpainting! Shorten it to only the face features, use low denoising strength, match the number of steps.

Recipe:

• iLustMix v1.0 (base checkpoint)

• Teeth XL 0.45

• Hand XL v5.5 0.45

• MoriiMee Gothic Niji Style Ilustrious Test 0.55

• Darker Than Dark Illustrious 0.55

• NoobAI-XL Detailer V-Pred1_V1 0.55

V2 - Alexx E. Edition

So, new stories, new checkpoint! This one is more of an experiment, i don't expect anyone to use it intensively, it does not behave well regarding following prompt :D

The base checkpoint used by DarkFun for the Alexx E. story is Illustrij V7, but i wanted to try some Block Merge, so i used as base here a "special" merge between this checkpoint and VauxzGrandMerge:

• BASE, MID: 0.5

• IN and OUT blocks from Illustrij: 0,2,4,7,8

• IN and OUT blocks from VauxzGrandMerge: 1,3,5,6

On top of this, i added the LoRA used at their respective weight:

• reij-detaILer: 0.3

• Incase_New_IllustV3: 0.25

• hand 5.5: 0.55

• illustrious_flat_color_v2: 0.8

For this one, Euler A + SGM Uniform is normally the way to go :D