How to Create Characters with Scenario: A Comprehensive Guide

Welcome to our detailed guide on creating consistent characters using Scenario, a revolutionary tool initially designed for game asset development that has found a wide array of applications in visual storytelling mediums like comics, visual novels, AI influencers, and more. Today we will learn how to create characters with scenario.

What is Scenario?

Scenario is not just a tool; it’s a game-changer for artists and developers alike. Originating as a game asset development tool, its primary strength lies in enabling the creation of characters that maintain consistency across various expressions, poses, and settings. This consistency is vital whether you are creating games, telling stories or creating characters to be used in various media.

Getting Started with Scenario

After signing up and logging into Scenario, You’ll be greeted with several options, including a “Start from Here” Section. We will ignore this and go straight to Generate, as the Start here options provide a very limited sandbox to try out the features to start generating we can click on Models on the left hand side panel.

A visual overview of the Main Page

Choosing Models

To start generating we need to select a base or starting model to start creating our character with. Scenario offers a wide range of models and assets, including curated models for various applications and community-contributed models, enriching the variety and possibilities for creation. Choose a model with a style that you like to get started.

The ‘Models’ section, showcasing the range of available models.

Creating a base for your Custom Character

Our current goal is to develop a prompt and generate batches of images off the prompt to get a handful of images that look consistent as the base for the character we want to create. I usually will generate between 20 -50 images and select 5-10 that look like they could be the same person.

A distinguishing characteristic of Scenario is its capability to produce a wide range of top-notch images. Through the utilization of the prompt assistant, individuals can refine their preferred art style, descriptors, aesthetics, and additional parameters to craft exclusive and visually captivating images. The prompt assistant offers an easy-to-use interface where users can choose from an array of options and explore various combinations.

Use the IP Adapter to help things along

Scenario has also recently implemented the IP adapter into the app. If you see a particular character that you like, you can download and use that image as the IP adapter reference to generate more images with the face of the character you want.

Building the LORA Dataset

Once we have selected a handful of images we like as the base for our character, we will use these to create a LORA. This will give us the maximum amount of flexibility with our character, and allow us to create multiple versions of them and combine them with other models.

To Train a Lora, we go to Models and then new Models on the top right.

Then we will get a Popup. We will select Train your own model

In the next step, we will upload the images we generated earlier. You will notice that Scenario will already provide a text description for each image. This is a great start, but we need to make some modifications to ensure our Lora creates a character we can reference.

To do this, we need to add a unique token or identifier to the text descriptions. Ideally you want to use a word that the model won’t confuse with something else. If you have a name for your character you can do something such as remove all the vowels. In my case I changed Lucy to LCY, as it’s unlikely the base model will have any other images encoded with that token. As you can see in the screenshot, I added this to each of the images.

Now it’s important to note, that each of the images I have uploaded are subtly different. However by using the same token, we will ideally be able to create a character that contains some consistent features across multiple generations. If necessary, you can reduce variance by repeating this process with images trained from the same model. A second “generation” model will ideally generate more consistency as we are converging the features of the character we are creating.

Before moving on, on the topic of captioning, although Scenario does a great job in captioning the training images, you want to take some time to go in and provide a good balance of details and descriptors. The better you can describe your image, the more flexibility you will have when using the model in dynamic situations. A Good Lora is a combination of a varied selection of images and good captioning. Below Is an image provided by Scenario to help showcase the difference in using a large dataset vs strong captions. Keep that in mind.

Once we are happy with our captioning we can go ahead and Start the training. One of the cool features of Scenario, is that depending on the number of images you decide to use, they will suggest some training parameters, that in most cases will be good to go. Regardless I have included some alternatives in the screenshot below

Examples of images within the LORA dataset showcasing the desired consistency.

Generating Images

Once the Lora has finished training, you will have the option to select it under model and generate images for it. Scenario has integrated tools such as Control net and IP adapter allowing you to have the maximum amount of flexibility in putting your character in all kinds of situations. Furthermore you can combine your Lora with other models to take your character into a multitude of art styles. Create a Character and then make their sprite version, give them emotions, and so much more!

If you want to try out scenario you can check it out here: https://app.scenario.com/?ref=endangeredai and don’t forget to use ENDANGEREDAI10 for a 10% discount

Empowering Creativity

In summary, Scenario goes beyond being just software; it fosters boundless creativity and innovation. By granting access to advanced design tools and resources, Scenario empowers creators of diverse backgrounds to explore their artistic potential freely. Positioned at the forefront of innovation, Scenario continues to drive the art of character creation with determination and ingenuity. From established characters like Serafina and Marina to the countless others yet to be imagined, Scenario opens the door to a new era of storytelling, unrestricted by imagination.