How to download NAI Diffusion and run the NovelAI model for free

An AI generated image of an anime girl against a neon-pixelated background

If you’re interested in learning how to download and run the same diffusion model that powers NovelAI Diffusion (NAI Diffusion 2), we’re going to walk you through it all.

While NovelAI is a paid service, being able to run the same diffusion model that generates its images will give you a lot more creative freedom, while also not being restrained by the same restrictions you might face online. The LLM is called animefull, which you can learn more about via Hugging Face.

From start to finish, the process won’t really take that much time, and it’s going to be especially worth it considering the benefits of having your own image-generation AI running from your home PC. While services such as DALL-E 3 and Photoshop generative fill are widely available – they’re also susceptible to outages and errors. Running the same model in NAI Diffusion and NovelAI will give you security against this. They also announced their new NAI Diffusion 2 model recently, with releases including NAI Diffusion Anime Curated, Full, and Furry.

Things to know before you install NAI Diffusion

Before you start downloading NAI Diffusion, there are a couple of things you might want to know. For example, it’s important to make a distinction between NAI Diffusion and NovelAI. The latter is a storytelling AI that turns plots and narratives into a digestible format. The diffusion model, on the other hand, will turn text-to-images with a pretty distinct Anime style.

You’re going to be running the animefull diffusion model from your computer, though this is going to require a user interface to work.

How to install the NAI Diffusion model

Before you attempt to install the NAI Diffusion model, you’re going to need to ensure you’re eclipsing the system requirements. For animefull, they are as follows:

  • 16GB RAM or above
  • Nvidia GTX 1070 with minimum 4GB VRAM, or AMD equivalent
  • Linux, Windows 7+ or Mac M1/M2
  • Minimum 10GB storage space

  1. Download the animefull model

    At some point last year, the NovelAI Diffusion model was leaked. The leakers turned the source code into a package that users could download – animefull – though it should be noted that it’s not as high quality as that of the original model.

  2. Install AUTOMATIC1111’s Stable Diffusion WebUI

    Installing and running Stable Diffusion locally lets you run diffusion models from your home PC, which is going to be how you’re going to get NovelAI Diffusion up and running.

    If you’re trying to run Stable Diffusion on Windows, you’re going to want to check out this article which runs over the install of AUTOMATIC1111. If you’re on Windows 10 or 11, it’s going to be as easy and downloading a .zip file, extracting it, then clicking the run.bat file to launch the AUTOMATIC1111 webui.

    Following this, you’re going to be given a local IP address which you’re going to be able to type into your browser’s URL bar. Paste it in and click Enter whenever you need to access the AUTOMATIC1111 Webui.

  3. Launch the animefull model through AUTOMATIC1111

    Locate the model.ckpt file alongside the config.yaml file and animevae.pt files. You’re going to place all three into the /stable-diffusion-webui/models/stable-diffusion/ folder.

    The main reason that we’re going to be renaming these files (we’ve attached those steps in the image) is that in the future, you could end up running multiple models via the same WebUI. Bespoke names will make it easier for you to adjust them in the future.

    After this, you’re going to run the WebUI by clicking on the .bat file. You can then access it by typing your local IP address into your web browser.A table displaying which files should be renamed.

  4. Test our your new NovelAI with the Hello Asuka prompt

    Positive Prompt masterpiece, best quality, masterpiece, asuka langley sitting cross legged on a chair

    Negative Prompt – lowres, bad anatomy, bad hands, text, error, missing fingers, extra digit, fewer digits, cropped, worst quality, low quality, normal quality, jpeg artifacts,signature, watermark, username, blurry, artist name

    Width / Height – 512 x 512
    Steps – 28
    Sampler – Euler
    CFG Scale – 12
    Seed – 2870305590

    If you enter the prompt as shown above, and it returns the right image, it’s going to mean that you have downloaded the NovelAI Diffusion model to your computer and you’re going to be able to run NovelAI from your PC.
    An image showing off the prompts needed to generate the ideal Hello Asuka image in NAI Diffusion.

Once you’ve finished the above steps, you’re going to have successfully installed the NovelAI diffusion model locally, meaning you’re not going to have to pay for the service.

Scroll to Top