coreml-ReV-Animated / README.md
Zabriskije's picture
Metadata update: we changed the `not-for-all-eyes` format (#1)
3bec478
metadata
license: creativeml-openrail-m
tags:
  - coreml
  - stable-diffusion
  - text-to-image
  - not-for-all-eyes

Core ML Converted Model:

  • This model was converted to Core ML for use on Apple Silicon devices. Conversion instructions can be found here.
  • Provide the model to an app such as Mochi Diffusion Github - Discord to generate images.
  • split_einsum version is compatible with all compute unit options including Neural Engine.
  • original version is only compatible with CPU & GPU option.
  • Custom resolution versions are tagged accordingly.
  • vae tagged files have a vae embedded into the model.
  • Descriptions are posted as-is from original model source. Not all features and/or results may be available in CoreML format.
  • This model was converted with vae-encoder for i2i.

Note: Some models do not have the unet split into chunks.

ReV Animated:

Source(s): CivitAI

Intro This model is mainly intended for Portraits and Full Body Anime-like pictures. Fantasy landscapes are decent.

READ MODEL VERSION DESCRIPTION ABOUT VAE THAT DIRECTLY AFFECTS STYLE RESULTS

No restrictions on use of this model, just credit to me and other model creators used in this merge.

Workflow:

Using AbyssOrangeMix2 recipe

if you would like to support me: https://ko-fi.com/s6yx0

About this version Last 4 photos are using orangemix.vae.pt

Base: A: Umi Olympus Landscapes v1.0 + B: Basil_mix_fixed = 1,0.9,0.7,0.5,0.3,0.1,1,1,1,1,1,1,0,0,0,0,0,0,0,0.1,0.3,0.5,0.7,0.9,1

Anime LoRA works real well with this model. Try them at 0.5-0.75.

Hires.fix:

Upscaler: ESRGAN_4x or ESRGAN_4x+ Anime6B or Remacri Upscaler

Upscale by: 1.5-2

Denoising strength: 0.4-0.5 (0.45 recommended)

Clip: 1

Vae:

orangemix.vae.pt - will give you more anime-like result, less 3D-like. More NSFW outputs, sfw if specified.

kl-f8-anime2.ckpt - will give you a more 3D type anime model, anime/cartoon aspect but more realistic. less NSFW unless specified.

Models used:

Umi AI Mythology and Babes by DutchAlex -> Great model series

Basil_mix_fixed

Cocoa NovelAI

Recommends

Resolution:

2:3 ratio (i.e 512x768)

1:1 ratio (i.e 512x512 or 576x576)

Best Samplers:

Euler a (Recommended)

DPM++ 2M Karras

CFG scale:

Portraits: 4.5-7

Landscape: 7-9

Steps:

Portraits: 30-50

Landscape: 30-50

Negative Prompts:

bad-xxx TI

i.e: (worst quality:1.4), (low quality:1.4), (monochrome:1.1), bad_prompt_version2, bad_artist_anime, (loli: 1.5), (shota:1.5), (child:1.4), ((disfigured)), ((bad art))

Prompts:

SFW: put 'NSFW' and 'uncensored' in negative prompt, SFW in regular prompt

NSFW: put 'SFW' and 'censored' in negative prompt, NSFW in regular prompt

Less negative prompts the better