Read Description

Questions/Feedback/Updates?

Visit my thread on the Unstable Diffusion Discord

Description

This model is intended to be used for merging purposes

Two trained models make up "Based" :

1. base01 - General purpose photorealistic model

2. HQSkin - Trained on highly detailed photos which featured clear skin details. Batch processed to maximize the fine details of the subjects skin.

Both models listed above are undertrained and are not available for download.

How to use the .yaml file :

For 'Based_v1.safetensors'

1. Simply drop the configuration file (.yaml) in the same directory as the model.

   1. Dir = stable-diffusion-webui\models\Stable-diffusion

If you're using another version of the model (such as 'Based_v1-FP16.safetensors')

1. Rename the .yaml file to match the name of the model.

   1. For example; For 'Base_v1-FP16.safetensors'. The .yaml should be renamed to 'Based_v1-FP16.yaml'.

2. Simply drop the configuration file (.yaml) in the same directory as the model.

Settings

VAE Required

Download VAE here

I prefer the .safetensors version, but the PyTorch (.ckpt) version is okay to use.

Place VAE inside :

stable-diffusion-webui\models\VAE

In webui :

Settings -> Stable Diffusion

Uncheck 'Ignore selected VAE for stable diffusion checkpoints that have their own .vae.pt next to them'

Webui Recommended Settings

1. ETA noise seed delta = 31337

   1. Settings -> Sampler Parameters -> Eta noise seed delta

2. Quick Settings = sd_model_checkpoint, sd_vae, CLIP_stop_at_last_layers, s_churn, always_discard_next_to_last_sigma

   1. Settings -> User interface -> Quicksettings

3. Hires Fix sampler selection : Enabled

   1. Settings -> User interface -> Hires fix: show hires sampler selection

Hires Fix

Models :

1. R-ESRGAN 4x+ | Denoise Strength = 0.3 - 0.35

2. 4x_RealisticRescaler_100000_G | Denoise Strength = 0.25 - 0.3

   1. Download here

   2. stable-diffusion-webui\models\ESRGAN

3. 4x_Valar_v1 | Denoise Strength = 0.5 - 0.6

   1. Download here

   2. stable-diffusion-webui\models\ESRGAN

Upscale Size <= 2

Sampler Settings

These are just recommendations, experiment with different settings.

1. DPM++ SDE Karras | 30 - 40 Steps

2. DPM++ 2M Karras | 30 - 60 Steps

3. Euler a | 20 - 40 Steps

4. DDIM | 80+ Steps

ADetailer

More info + install instructions here

1. Enabled = True

2. Model = face_yolo8n.pt (experiment with other models from dropdown)

3. Detection model confidence threshold = 0.3 - 0.8 (experiment with your own settings)

CFG

Mainly tested for DPM++ SDE Karras

This model seems to like work better with values above 7. Typically I found myself using 7.5 or 8.5 for most gens.

Note : If using Euler a, reduce CFG from ~6 - 7

Check out my other models

SDXL

• Boomer Art Model - https://civarchive.com/models/163139/boomer-art-model-bam

• LomoXL - https://civarchive.com/models/156743/lomoxl

SD1.5

• Doomer Boomer - https://civarchive.com/models/118247?modelVersionId=128239

• Lomostyle - https://civarchive.com/models/109923/lomostyle

• Electric Eden - https://civarchive.com/models/64355/electric-eden

• Cine Diffusion - https://civarchive.com/models/50000/cine-diffusion

• ProjectAIO - https://civarchive.com/models/18428/project-aio

• Eris - https://civarchive.com/models/21952/eris

• WonderMix - https://civarchive.com/models/15666/wondermix

• Refined - https://civarchive.com/models/8392/refined

• Experience - https://civarchive.com/models/5952/experience

• Elegance - https://civarchive.com/models/5564/elegance

• Clarity - https://civarchive.com/models/5062/clarity

• VisionGen - Realism Reborn -https://civarchive.com/models/4834/visiongen-realism

LoRA

• Pant Pull Down - https://civarchive.com/models/11126/pant-pull-down-lora