---
license: creativeml-openrail-m
language:
- en
library_name: diffusers
pipeline_tag: text-to-image
tags:
- text-to-image
- stable-diffusion
- diffusers
- safetensors
- stable-diffusion
- anime
- style
base_model: runwayml/stable-diffusion-v1-5
---

# Nabla Theta A5 - Lines and colors enhanced A5 

## Notes:
- Original Author: [`AD_KOISHI`](https://civitai.com/user/AD_KOISHI)
- Main Checkpoint: https://civitai.com/models/260571/nabla-theta-a5-lines-and-colors-enhanced-a5-a5
- Restructured by: [`Eien Mojiki`](https://huggingface.co/moji-ai)

## Model Introduction
- **NablaThetaA5-v1.0** is an anime-style model fine-tuned from **SD-v1.5**. This is a restructured model suitable for use with the **`Diffusers`** library.
- **Some examples**:
![#](https://i.postimg.cc/d0vDrSGn/Comfy-UI-00001-2.png)
![#](https://i.postimg.cc/fTJdRFv9/Comfy-UI-00029.png)
![#](https://i.postimg.cc/Kz8zskqM/Comfy-UI-00030.png)

## Usage
1. **Using Diffusers**

Make sure you have installed Diffusers and necessary libraries
```shell
pip install --upgrade diffusers accelerate transformers
```
Quick example:
```python
import torch
from diffusers import StableDiffusionPipeline, EulerAncestralDiscreteScheduler

pipe = StableDiffusionPipeline.from_pretrained("moji-ai/NablaThetaA5-v1.0", torch_dtype=torch.float16).to("cuda")
pipe.scheduler = EulerDiscreteScheduler.from_config(pipeline.scheduler.config)

prompt = "some cool prompt here"
image = pipe(prompt).images[0]
```
2. **Use in A1111 or ComfyUI**
- [A1111](https://github.com/AUTOMATIC1111/stable-diffusion-webui): Download the model [here](./NablaThetaA5-v1.0.safetensors) > Place the model inside the `models/Stable-diffusion` directory of your AUTOMATIC1111 Web UI instance > Refresh your model list or restart the Stable Diffusion Web UI > Select the custom model from the Stable Diffusion checkpoint input field and make awesome images!!
- [ComfyUI](https://github.com/comfyanonymous/ComfyUI): Download the model > place it inside `ComfyUI/models/checkpoints/` directory > Restart ComfyUI and make cool images!!

**Recommended settings**: 
- `CFG`: 5.5~7
- `Sampler`: Euler a
- `Steps`: 28~35
- `High-res`: [RealESRGAN_x4plus_anime_6B](https://github.com/xinntao/Real-ESRGAN/blob/master/docs/anime_model.md) / 15 steps; 0.55~0.6 denoising strength

> It is recommended to use the [A5 Stabilizer](https://civitai.com/models/255938/a5-stabilizer-rlhf-d3po-a5) LORA for generating to see slightly different styles under the same random seed, and in some cases, improve the image to better match your preferences.

## From Author:
- The author strongly dislikes many homogenized styles nowadays, so the author trained and shared this model.

***This is an incomplete model card, so feel free to open a discussion to improve this model***