Capitan's Illustrious

V2.2 Update:

Check on "About this version" for changes.

Positive keys and order of prompting suggested:

[perfect quality, detailed background, intricate details, highres, hi res, best quality, masterpiece, anime coloring]

Amount of subjects / Angle and type of picture / The principal idea / details / background / accents

Examples:
1- 1boy, 2girls, side view, high angle, upper body, young people at party, two girls in dress drinking beer with a boy, in living room, balloons and food on tables

2- 1boy, low angle, POV aiming at male, mature male looking down at viewer scared, aiming gun at mature male, outdoors, park at night

Negative keys suggested:

lowres, worst quality, bad quality, bad anatomy, (bad hands), username, logo, watermark, signed, signature, sketch, watermark, artist name, censored, low quality, blurry, distorted, patreon, <Capitan_error>

Samplers (from better to regular):

Euler_A, Euler, DPMpp 2M, DPMpp SDE

Schedulers (from better to regular):

Simple, Normal, Karras

16 ~ 34 steps

4.5 ~ 7 cfg
__________________________________________________________
V2 Update:

Check on "About this version" for changes on CLIP and prompt guidance.
__________________________________________________________
V1.6 Update:

Thanks for the buzz donators and Ko-Fi tips, I wanted to show my gratitude making this version.

At the start of your prompts:

/Masterpiece, best quality/

At the end of your prompts:

/ High quality, CapSer, Very detailed/

!!!

Sometimes you might get ride of "CapSer" since it can add more contrast and lighting details but also force characters on "no humans" compositions, due to pictures with this tag having always 1 or more humans, monster or animals.

Negative:

bad quality, worst quality, worst detail, artifacts, signed, sketch, censor, (bad hand, bad feet, long chest), (moe themed, chibi, cute themed)

_______________________________________________________________________________

Esto es un experimento que me llevó dos días, me gustó como quedó y pues aquí va.

El dataset usado es el mismo que usé para PonyMan Hyper v2.5 [ https://civitai.com/models/2271725?modelVersionId=2668729 ]

¬ Steps from 17 to 35. More than 36 tend to add artifacts and supress LoRas

¬ CFG scale should be from 4.5 to 7.5 (hard limit, since it starts adding artifacts or burning the skin)

¬ Samplers Euler Ancestral for fastest process, DPM++ 2M Karras for better quality (slower)

¬ Scheduler: normal of sgm_uniform

The VAE is already baked in.

Conditions of use:

1- You are the only responsible for your creations.

2- If you were to sell the pictures, I'm not asking for money or shit, just mention the author of this (me).