Baka-Diffusion - A latent diffusion model fine-tuned to output High Quality anime illustrations!**
________
# Baka Diffusion V1
Welcome to Baka-Diffusion! a latent diffusion model Trained and Fine-Tuned on **High Quality** images using the **Danbooru** tagging dataset! our models are made to output better lighting and quality with just a few tags!
e.g. **_masterpiece, best quality, 1girl, hakurei reimu, chromatic aberration, white background,_**
________
# **The model was made to use with Latent Space Upscaling in mind**
*Below are some examples generated with this model!*
```
masterpiece, best quality, 1girl, hakurei reimu, chromatic aberration, white background,
Negative prompt: lowres, (bad anatomy), text, cropped, worst quality, low quality, jpeg artifacts, signature, watermark, username, artist name, (out of frame), black and white, obese
```
# Versions!
Our models have 2 versions FP16 and FP32 from my testing FP32 can create more complex backgrounds where as for the FP16 it creates simpler backgrounds
# Usage!
- DPM++ SDE Karras seems to be the best sampler for BakaV1
- ClipSkip 1
- Upscaler (Latent)
- Denoise Strength (0.5~0.7)
- below 0.5 Denoise Blocking can occur
## Changelog!
- Removed Ckpt versions 1/16/23
- Added Diffuser weights! 1/16/23
- Model page overhaul 1/16/23
## This Project would be impossible without
- [HaisenBerg](https://huggingface.co./haisenberguwu)
# Baka-Diffusion Version2
Baka-DiffusionV2 has been delayed :( I have been working on the model for weeks and have come to a dead end, There were working test versions but the quality isn't as notable as i thought it would be, It felt like a Version1.5 than a V2 so i have restarted my entire procress but eventually it will come out but I cannot pin point an exact date so please be patient!
Many thanks,
Hosioka.
-----------------------------------------------------------------------------------------------------------------------------------------------------------------------------
This model is open access and available to all, with a CreativeML OpenRAIL-M license further specifying rights and usage.
The CreativeML OpenRAIL License specifies:
1. You can't use the **Model** to deliberately produce nor share illegal or harmful outputs or content
2. The authors claims no rights on the outputs you generate, you are free to use them and are accountable for their use which must not go against the provisions set in the license
3. You may re-distribute the weights and use the **Model** commercially and/or as a service. If you do, please be aware you have to include the same use restrictions as the ones in the license and share a copy of the CreativeML OpenRAIL-M to all your users (please read the license entirely and carefully)
[Please read the full license here](https://huggingface.co./spaces/CompVis/stable-diffusion-license)