Sign In

Kindred

Download

2 variants available

Type

Checkpoint Merge

Stats

417

Reviews

Published

Jul 18, 2023

Base Model

SD 1.5

Hash

AutoV2
778EA98D2B
Comfy Workflow Badge
sca285's Avatar

sca285

Overview

If you're looking to dive into the world of Niji-like models, get ready to unleash your creativity and immerse yourself in Kindred. This model was inspired by Nijiv5 style, can typically be used for various creations such as characters, styles, background, fantasy and sci-fi themes, and NSFW. This is a mix of merging and self-trained efforts

If you love this one, please give some likes, reviews with pictures, and 5 stars ^_^

Recommended Textual Inversion Embeddings

EasyNegative, Simplenegative, deformityv6, bad-hands-5, UnrealisticDream, BadDream, bad-artist-anime


Recommended Settings

  • Size: 512x768 (default)

  • Sampling method: DPM++ 2M Karras, DPM++ 2M alt Karras, DPM++ SDE Karras, DPM++ 2M SDE Karras,

  • Hires.fix upscaler: R-ESRGAN 4x+ Anime6B, NMKD-Yandere variations, Remacri

  • CFG Scale: 7~9

  • VAE : vae-ft-mse-840000-ema-pruned, klF8Anime2VAE, orangemix.


Log

V1.0 : Double release Bold and Soft. Bold inclines on a 2.5D semi-real with darker tones while Soft does anime-like style with pastel tones

  • This model requires you to put exactly the type of pictures you wanna create so you have to put the correct prompts such as portrait/cowboy shot/upper body/landscape... etc.

  • The human-shape body is default, but I haven't tried a humanoid-body creations (furry of course), wonder how capable it could do

  • Different VAE, LORA, embeddings give different results, you may not get the original face


Merge Material

Models:

AnythingQingMix-2.5D

Glow 2.5D

BreakDomain

Dalcefo

LORAs:

Detail Tweaker LoRA

deluster2

Brightness Tweaker LoRA

yawanuri