|
--- |
|
license: creativeml-openrail-m |
|
library_name: diffusers |
|
tags: |
|
- stable-diffusion |
|
- text-to-image |
|
--- |
|
|
|
# Capabilities |
|
|
|
This model is "adventure" and "fantasy" focused. |
|
|
|
With certain inference configurations, it is capable of producing very high quality results. |
|
|
|
This model functions better without negative prompts than most fine-tunes. |
|
|
|
# Inference parameters |
|
|
|
Diffusers should "Just Work" with the config in this repository. |
|
|
|
For A1111 users, |
|
|
|
Scheduler: DDIM, 15-50 steps |
|
Generally acceptable resolutions: |
|
- 768x768 |
|
- 1024x1024 |
|
- 1152x768 |
|
|
|
# Limitations |
|
|
|
This model contains a heavily tuned text encoder that has lost many original Stable Diffusion 2.1 concepts |
|
|
|
This model is even less reliable at producing real people than the base 2.1-v model is |
|
|
|
Training data included only 768x768 downsampled 1:1 ratio images, all other aspects were discarded. Ergo, this model struggles with high resolution native generations. |
|
|
|
This model may have "burnt" outputs at higher CFG. |
