Colorful-v4.5-Plus / README.md
Narcis
Update README.md
3cf5f0e
|
raw
history blame
3.91 kB
metadata
license: creativeml-openrail-m
language:
  - en
library_name: diffusers
pipeline_tag: text-to-image
tags:
  - stable diffusion
  - text-to-image
  - diffusers

Colorful-v4.5

Colorful-v4.5 is a model merge between Anything-v4.5, AbyssOrangeMix2 and ProtogenInfinity

Colorful-v4.5 is named what it is because of the fact that it is similar to Anything-v4.5 and that it improves the bland color pallet it comes with (atleast for me), producing much livelier images. It also improves some other things like environments, fingers, facial emotions and somewhat clothing (it also fixes the purple blobs 🤫)

technically i could call it Anything-v5.0 but that would be rather cheesy and rude

Though it is note-worthy that it is a bit more unstable in some areas (Example #2)

Attention: It is highly recommended to run this model locally on your computer because running it from the web-ui api will produce lower quality images than intended

Examples:

Example #1: 10 steps

02040-774768794-masterpiece, best quality, 1girl, (green eyes), black hair, (black shorts), white shirt, glossy lips, small nose, standing up, i.png

Prompt: masterpiece, best quality, 1girl, (green eyes), black hair, (black shorts), white shirt, glossy lips, small nose, standing up, in park, trees, full body, smiling
Other Details: Steps: 10, Sampler: DPM++ 2S a Karras, CFG scale: 8, Seed: 774768794, Size: 512x512, Model hash: b5de490700, Model: Colorful-v4.5, Denoising strength: 0.6, Hires upscale: 2, Hires steps: 10, Hires upscaler: SwinIR_4x
Negative Prompt: The negative prompt is very long and specific so it will be listed in the model's repo. ( The negative prompt comes from another model called Hentai Difussion so it will contain NSFW. A curated version of the negative prompt will also be in the repo for those who want SFW)

Example #2: 20 steps

02041-774768794-masterpiece, best quality, lady, in red and black yukata, pink hair, blue eyes, in dojo, smiling, sitting, hands on lap.png

Prompt: masterpiece, best quality, lady, in red and black yukata, pink hair, blue eyes, in dojo, smiling, sitting, hands on lap
Other Details: Steps: 20, Sampler: DPM++ 2S a Karras, CFG scale: 8, Seed: 774768794, Size: 512x512, Model hash: b5de490700, Model: Colorful-v4.5, Denoising strength: 0.6, Hires upscale: 2, Hires steps: 20, Hires upscaler: SwinIR_4x
Negative Prompt: Same thing as in Example #1

Example #3: 30 steps

02048-774768794-masterpiece, best quality, girl, black hair, blue eyes, black t-shirt, black pants, smiling, standing up, solo, facing viewer, n.png

Anything-v4.5: (for comparison)

02049-774768794-masterpiece, best quality, girl, black hair, blue eyes, black t-shirt, black pants, smiling, standing up, solo, facing viewer, n.png

Prompt: masterpiece, best quality, girl, black hair, blue eyes, black t-shirt, black pants, smiling, standing up, solo, facing viewer, near blossomed tree
Other Details: Steps: 30, Sampler: DPM++ 2S a Karras, CFG scale: 8, Seed: 774768794, Size: 512x512, Model hash: b5de490700, Model: Colorful-v4.5, Denoising strength: 0.6, Hires upscale: 2, Hires steps: 30, Hires upscaler: SwinIR_4x
Negative Prompt: Same thing as in Example #1

Note: I didnt use any vae for the examples, but i did try the anything-v4.0 vae and it barely made a difference