Hugging Face's logo

AshtakaOOf
/
Amedira

Text-to-Image
Safetensors
English
art
anime
stable-diffusion
Model card Files Community
Amedira / README.md
Adrien
Update README.md
b06ac47
preview code
|
raw
history blame
6.26 kB
metadata
license: creativeml-openrail-m
language:
  - en
pipeline_tag: text-to-image
tags:
  - art
  - anime
  - stable-diffusion
library_name: safetensors

🎑 ~ Amedira ~ 🎏

🎈 ❱ A model merge using classic merging and clip transfer

This model is capable of doing multiples styles, such as flat and 2.5d.

You can also prompt Hololive vtubers with their danbooru tags (you will need to prompt their outfit too for some of them).

And it is somewhat compatible with LoRA and lyCORIS.

⚠️ Don't forget to use the following token:

  • SLE is a token that will improve eyes quality and saturation, and it come from AuroraOne.
SLE

Note: Don't change the weight of this token and keep it at 1.0

Other style tokens:

  • Semi Realistic
(realistic:1.2), (mksks style, detailed background)
  • Flat style:
(flat color:1.2), (pastel_style, anime, black_outlines)

Note: These tokens won't work very well on the Dreams variant

📥 ❱ Downloads

Amedira

Dreams

Note: The Dreams variant are experimental/accidental merges that I keep because they have unique styles

Lunamedira

Note: These checkpoints are hard to use

🖼️ ❱ Images Comparisons

🧬 ❱ Negative Box

Negative Embeddings combinations:

  1. KHFB, AuroraNegative, negative_hand-neg
  2. verybadimagenegative_v1.3, negative_hand-neg
  3. verybadimagenegative_v1.3, simple5.5, simple5.6

Use these to complement the Negative Prompts by putting them before it

Negative prompts:

Efficient

[worst|low] quality, nsfw, lowres, bad anatomy, watermark

Classic

(low quality, worst quality:1.4), nsfw, bad anatomy, bad hands, artist name, watermark

Simplest

(worst quality, low quality:1.4), nsfw

🥁 ❱ Credits