Streamline the Creation of NFT Collections with AI

NFT collection screenshot

Overview

Creating an entire NFT art collection can be time-consuming and challenging. However, with AI, this process becomes effortless. It's never been easier!

Three alien variants

In this guide, you'll learn how to use Scenario to quickly and easily design a captivating character and iterate on it to create a potential NFT collection.

Getting Started

  1. Training a Consistent Model
  2. Creating the First Edition Character Variations

Step 1: Training a Consistent Model

To start, it is important to understand the principles of training a consistent character model, so you can train your own NFT model in the future. For this workflow we have created training images, however you can follow the same approach with your own training images of a consistent character.

For a consistent character training dataset you should have 8-15 images of your consistent character in various poses and angles, using different props or expressing different emotions.

Download Training Images

Choose a custom token for your character and modify your captions to include that token. For the character we are using named ‘snugglix’, simply remove the vowels to create a unique token “sngglx”. Add this token to each caption.

Make note of other descriptors in the captions that occur often as you will need to remember to use those caption descriptions in your prompts to later invoke the character on your custom model.

In this instance, the caption often reads “sngglx is wearing a jacket”, so use this prompt on the model later if you want Snugglix to be wearing his gold jacket. Select the ‘Subject’ training default and Train the Model. Use the premade captions linked below.

Captions for the Training Images

Now select the Subject training preset. Do not make any changes to Training Parameters.

Examples of good captions

Step 2: Creating the First Edition Character Variations

Basic Prompting

Begin prompting your new custom character model and create an image to iterate on. Remember to use your unique token! For the examples here, we used the following prompt on the Snugglix model:

sngglx wearing a jacket

You can also add objects or costume elements by prompting for them to add more variation to the collection. In this way, you can quickly and easily create a full edition of your collection. Here are the prompts we used to create a couple more base images.

sngglx holding a guitar

sngglx in a space suit, wearing a space helmet

Multi Reference Image Prompting

Drag and drop the image to the Reference Image tab, and then select the mode Image to Image + Style Reference.

Use your generated character image for Image to Image, and select a separate image with a color palette you would like to use to reskin the character for Style Reference. Remember to maintain the original prompt you used to generate your Image to Image reference in the prompt window for the most consistent results. Set the Style Reference Influence to 50 and the Image to Image Influence to 25.

A style transfer example

Create as many reskins as desired by keeping the Image to Image reference the same and replacing the Style Reference reference.

Reskin examples

Conclusion

Using this guide, you can quickly and easily design a full NFT collection from start to finish. Generate a wide array of unique character designs, then hone in on your favorites and train custom models to create the core of your collection. This streamlined process makes it simple to constantly refresh and expand your NFT offerings, ensuring you can keep your community engaged and your collectors coming back for more.

Additional Links

More information on IP Adapter

More information on curating and captioning your dataset

Table of content

You'll also like