Verified: 2 years ago
SafeTensor
Type | |
Stats | 5,163 42,260 |
Reviews | |
Published | Aug 2, 2023 |
Base Model | |
Training | Steps: 7,000,000 Epochs: 100 |
Usage Tips | Clip Skip: 2 |
Hash | AutoV2 7B7F6EE6FB |
First release of Radiance merged with MeinaHentaiV4. It still aims at anime feeling with vibrant color, which will stay on the Radiance series. Prompting is hard as it is not a merge model like you are used to see. Not recommended for beginners.
After dowloading rename the file to : AstreaPixie_RadianceHentai_v1.6pruned
This is because Civitai limits model name to 20 characters
Recommended Settings
- Clip skip: 2
- CFG scale: 7
- Sampler: Euler, Euler A, DPM++ SDE and DPM++ 2m Karras
- Sampling Steps : 20-30
- Upscaler (Hires. fix): 4x_NMKD-Superscale-SP_178000_G
- Hires steps: 10-15 Denoising strength: 0.4 Upscale by: 1.5+
Using Adetailer
- Main UI > Extensions > Install from URL
This mainly fixes problems with faces and eyes. Do not use for hands.
This can be enable and nothing else to do to works.
Why is my results different from yours?
Enable quantization in K samplers in your settings
Keep your prompts simple and correct
Tag auto complete extension to help you with your tags
Neg embedding have a great impact on the output of a model
Depending on the one you are using, results will be different.
I mainly use fitCorder embeddings for negative :)