How to Add the Ӿ Symbol to Your AI Image Generations

One of the more recent technological revolutions has been in AI, with image generation proving to be a very popular and rapidly advancing sub-category. You may have tried generating these images yourself through tools like NanoGPT. The results are often very impressive and sometimes a little weird, but always interesting.

A commonly-voiced frustration within the Nano community is not being able to get the Nano currency symbol (Ӿ) into these image generations. Normally a compromise is reached by generating an image and photoshopping the Ӿ symbol onto it afterwards, which can look good but sometimes a little flat.

Why is it difficult?

The reason it is difficult to get the Ӿ symbol into image generations is that there is no easy way to describe the Ӿ symbol to the base AI model. An AI model will have knowledge based upon images and captions it has been trained with, so if the model has not been fed any information about the Ӿ symbol it will not understand when you try to describe it.

We need a way to get our image generation model to understand the Ӿ symbol. This article will describe one way to do this today, locally on your computer.

Hardware requirements

To perform this you will need a computer with:

• Windows installed (Tested on a Windows 10 setup. May work on Linux/Mac also but not tested)
• An Nvidia graphics card with a minimum of 6GB VRAM
• Recommend minimum of 16GB RAM

Later in the article we will mention an alternative method if your computer does not meet these requirements.

Software requirements

Download and install the open-source image generation software Fooocus. During installation it will also download the AI model necessary to generate your images. Note that the model is quite a large file (approx. 6GB).

Once you have installed Fooocus you will need to download a smaller model that contains knowledge about the Ӿ symbol. Download the Ӿ symbol AI model (218 MB) and place it into the folder:

<your_fooocus_install_directory>/Fooocus/models/loras

Open up Fooocus and you will be provided with a minimal interface with which to generate images using prompts.

Before you begin

First make sure the Advanced checkbox is checked.

The advanced options should become visible. Click the Model tab. Under LoRA 2 select xno_symbol-SDXL.safetensors. This "plugs" the Ӿ symbol model into your base model. Think of it like a small Lego brick being added to a larger Lego structure.

Let's get generating!

You are now ready to start generating your images. Click inside the Type prompt here input box and enter, for example:

xno_symbol on a cat's hat

What is the "xno_symbol" text in the prompt? This is the keyword that the AI model understands as the Ӿ symbol. It must appear exactly as shown, not "xno", not "nano_symbol" - it must be "xno_symbol". The keyword can be written anywhere in your prompt, not just at the beginning.

After you have entered the prompt, hit Generate.

Depending on the power of your computer you should see your image generation after a few seconds. Here is the generation we got:

To create different styles of image you can use the Style tab to select from a multitude of preset styles.

 

More Ӿ symbol image generations

Prompt: wind turbine, standing on mountaintop, giant xno_symbol painted in blue on ground below
Style: Ads Corporate

 

Prompt: woman with xno_symbol jacket in city
Style: Neonpunk, Anime

 

Prompt: people looking at electricity xno_symbol
Style: Artstyle Hyperrealism

 

Prompt: cute broccoli dog in kennel with xno_symbol
Style: Craft Clay

 

Prompt: giant xno_symbol statue standing in a beautiful meadow, isometric view
Style: Comic

 

Prompt: red robot with 3d xno_symbol on chest, factory
Style: Futuristic Cybernetic Robot

 

Prompt: person holding a smartphone, xno_symbol on screen, close-up, busy street
Style: Ads Corporate

 

My Ӿ looks weird

From time to time you will likely get some wonky-looking Ӿ symbol generations (e.g three horizontal lines, instead of two). The AI will not get it perfect every time and this is likely because the Ӿ symbol model was not trained on enough data. There are ways to fix weird generations but they are outside the scope of this article. Hint - try experimenting with Fooocus' Inpaint feature.

My computer doesn’t meet the hardware requirements

All is not lost. There are services online that provide cloud instances of Fooocus for a small cost and should allow you to plug in the Ӿ symbol model to generate images like we did here. ThinkDiffusion is one example of such a service. We didn't get around to testing any of these cloud services so your results may vary.

How can it get easier in future?

Other image generation tools should gradually start to understand what the Ӿ symbol is when they encounter it in the wild through their training data. So that's another good reason to start using the Ӿ symbol more in your graphics and photos. Let’s get it out there! To obtain a clean, perfectly formed version of the Ӿ symbol for your graphics head on over to our useful Assets page.