README.md

---
license: creativeml-openrail-m
language:
- en
library_name: diffusers
tags:
- pytorch
- art
- anime
- text-to-image
---

# Kawai Diffusion (anime-base) v3.0 LTS Big Update (≧∇≦)ﾉ
**Waifumake from now on will be Kawai Diffusion.**

See more in CivitAI : https://civitai.com/models/21138/kawai-diffusion-sd15
![](asset/preview.png)
## What's new in Kawai v3.0 LTS:
- Fix color loss.
- Image quality is greatly enhanced. Thank you my friend.
- Kawai Diffusion's most powerful ability is "enhance" (img2img). It will make a bad photo look better.
- True "kawaii"... Haizzzzzzzzz
- Two versions: the [ema-only](kawai-diffusion-sdv15_ema-only.safetensors) model (5.28GB), and [pruned model](kawai-base-sd-v15_pruned-full.safetensors) (8.49GB). Come on, don't be surprised by it, even I was surprised.
- Can work on some VAE. But the pruned model does not require any VAE.

## Introduction:
- It's an AI art model for converting text to images, images to images, inpainting, and outpainting using Stable Diffusion.
- The AI art model is developed with a focus on the ability to draw anime characters relatively well through fine-tuning using Dreambooth.
- It can be used as a tool for upscaling or rendering anime-style images from 3D modeling software (Blender).
- Create an image from a sketch you created from a pure drawing program. (MS Paint)
- The model is aimed at everyone and has limitless usage potential.

## Use:
- For 🧨Diffusers:
```python
from diffusers import DiffusionPipeline

pipe = DiffusionPipeline.from_pretrained("Ojimi/anime-kawai-diffusion")
pipe = pipe.to("cuda")

prompt = "1girl, animal ears, long hair, solo, cat ears, choker, bare shoulders, red eyes, fang, looking at viewer, animal ear fluff, upper body, black hair, blush, closed mouth, off shoulder, bangs, bow, collarbone"
image = pipe(prompt, negative_prompt="lowres, bad anatomy").images[0]
```
- Try it in Google Colab [![Open In Colab](https://colab.research.google.com/assets/colab-badge.svg)](https://colab.research.google.com/drive/1D6LNtXrpD2QfUx-d_yztWZVgTiDAyyAT?usp=sharing)
- Chat GPT with Kawai Diffusion (or any model if you like.)

```code 
Read the following instructions, and if you understand, say "I understand": Command prompt structure: includes descriptions of shape, perspective, posture, and landscape,... Keywords are written briefly in the form of tags. For example "1girl, blonde hair, sitting, dress, red eyes, small breasts, star, night sky, moon"
```
## Tips:
- The `masterpiece` and `best quality` tags are not necessary, as it sometimes leads to contradictory results, but if it is distorted or discolored, add them now.
- The CGF scale should be 7.5 and the step count 28 for the best quality and best performance.
- Use a sample photo for your idea. `Interrogate DeepBooru` and change the prompts to suit what you want.
- You should use it as a supportive tool for creating works of art, and not rely on it completely.
- The Clip skip should be 2.

## **Training**:
- **Data**: Created by another AI.
- **Schedule**: DDIM.
- **Optimizer**: AdamW.
- **Precision**: FP32.
- **Hardware**: Google Colaboratory Pro - NVIDIA A100 40GB VRAM, TESLA V100-SXM2 16GB.
## **Model Unit Test:**
This is a program written by my friend to check model quality.
- Examiner: OpenAI ChatGPT-3.5-Turbo.
- Test: kawai-anime-sd.
- Schedule: DPM++ 2M Karras.
- Steps: 22.
- Guard: Guard Prompt 1.5.
- Test Report: [Here](https://civitai.com/gallery/299771?modelId=21138&modelVersionId=27219&infinite=false&returnUrl=%2Fmodels%2F21138%2Fkawai-diffusion-sd15).

## **Limitations:**
- The drawing is hard, not soft.
- Loss of detail, errors, bad human-like (six-fingered hand) details, deformation, blurring, and unclear images are inevitable.
- ⚠️Content may not be appropriate for all ages: As it is trained on data that includes adult content, the generated images may contain content not suitable for children (depending on your country there will be a specific regulation about it). If you do not want to appear adult content, make sure you have additional safety measures in place, such as adding "nsfw" to the negative prompt.
- The results generated by the model are considered impressive. But unfortunately, currently, it only supports the English language, to use multilingual, consider using third-party translation programs.
- The model is trained on the `Danbooru` and `Nai` tagging system, so the long text may result in poor results.
- My amount of money: 0 USD =((. 

![](money-wallet.gif)

## **Desires:**
As it is a version made only by myself and my small associates, the model will not be perfect and may differ from what people expect. Any contributions from everyone will be respected.

Want to support me? Thank you, please help me make it better. ❤️

## Special Thank:
This wouldn't have happened if they hadn't made a breakthrough.
- [Runwayml](https://huggingface.co/runwayml/):  Base model.
- [d8ahazard](https://github.com/d8ahazard/.sd_dreambooth_extension) : Dreambooth.
- [Automatic1111](https://github.com/AUTOMATIC1111/) : Web UI.
- [Mikubill](https://github.com/Mikubill/): Where my ideas started.
- Chat-GPT: Help me do crazy things that I thought I would never do.
- Novel AI, Anything Model, Abyss Orange Model: Dataset images. An AI made me thousands of pictures without worrying about copyright or dispute.
- Danbooru: Help me write the correct tag.
- My friend and others: Get quality images.
- And You 🫵❤️

## Copyright:

This license allows anyone to copy, and modify the model, but please follow the terms of the CreativeML Open RAIL-M. You can learn more about the CreativeML Open RAIL-M [here](https://huggingface.co/spaces/CompVis/stable-diffusion-license).

If any part of the model does not comply with the terms of the GNU General Public License, the copyright and other rights of the model will still be valid.

All AI-generated images are yours, you can do whatever you want, but please obey the laws of your country. We will not be responsible for any problems you cause.

We allow you to merge with another model, but if you share that merge model, don't forget to add me to the credits.

Don't forget me.

# Have fun with your waifu! (●'◡'●)

I have a hero, but I can't say his name and we've never met. But he was the one who laid the foundation for Kawai Diffusion. Although the model is not very popular, I love that hero very much. Thank you for your interest in my model. Thank you very much!

Like it? Buy me ko-fi: https://ko-fi.com/ojimi (≧∇≦)ﾉ