Type | |
Stats | 741 |
Reviews | (108) |
Published | Apr 12, 2023 |
Base Model | |
Hash | AutoV2 1FAF8EB819 |
Filmation's Masters of the Universe Style
Features a wide variety: humans, creatures, dragons, aliens, landscapes, forests, and architecture in the specific style of Filmation's Masters of the Universe. I grew up watching this show and being mystified by the story line around Castle Grayskull. Enjoy!
> 🌟 This LoRA in an upgrade from the original Dreambooth checkpoint
Improvements over CKPT version
File size: 36.1MB vs 3.82GB
Text encoder trained on DeepBorru tags: https://danbooru.donmai.us/tags
Additional training input: dragons, tigers, landscapes, architecture, emotion, clothing - see gallery for prompt examples.
Model Usage
Trained using kohya-ss SD scripts.
Using the Additional Networks extension within Auto1111 is recommended as you can adjust UNet and Text Encoder weights independently.
Cartoon Style
Use any NAI or animation/cartoon type model trained with booru tagging:
Weight: 0.1
- 0.9
Clip Skip: 2
> 💥 VAE Check: Double check to make sure you're using the correct VAE per model. Having a VAE active (especially the correct one) will give you the best results.
Realistic Style
The style can transfer as LoRA weights to other models. Below are a few that I used to generate the sample images.
Weight: 0.1
- 0.55
Upscaler: R-ESRGAN 4x+
Denoising strength: 0.5
- 0.63
img2img Settings
This loRA was extensively tested with a large variety of img2img results. See various img2img samples in the Gallery.
Denoising Strength: 0.58
CFG: 8
Steps: 20
LoRA Weight: 0.7