Workflow for Generating Multi-Character Scenes

Introduction

In this article, we present a workflow that allows you to leverage your trained character models and styles to quickly iterate and construct captivating scenes featuring multiple characters, revolutionizing the way you create immersive gaming experiences. From multi character scenes, marketing materials, and even web toons, the multi character workflow unlocks a lot of possibilities.

Overview

In the following sections, we will guide you through:

Training and composing your character and style models.
Generating your base image.
Creating your characters.
Rapidly assembling the scene in Canvas.

If this is your first project with Scenario, we recommend approaching each step sequentially for the smoothest experience.

‍

‍Step 1: Training and composing your character and style models.

The most crucial decision in this workflow is to define who your characters are. The beauty of this process lies in its ability to accommodate multiple distinct characters, allowing you to achieve great results, no matter the number of characters you choose to include.

If you are working from references, it is possible your character styles vary slightly. It’s wise to train a Style Model. Doing so can significantly enhance the overall coherence of your scene, ensuring that your chosen style complements your character designs, and locking the characters into a consistent design language.

You will need at least two character models and one style model for this workflow. Make sure to create a separate Composition for each character with your style model for best results, and one model that combines both character models and the style. For comprehensive instructions on training your character and style models, as well as tips on composing them for optimal results, click the link below.

Training a Character Model

Training a Style Model

Create a Composition

‍

Step 2: Generating your base image

With your models prepared, it's time to generate your base image. This step is crucial for establishing the ideal design foundation. You can create your base image directly from the style model you're using or use a reference image sourced from elsewhere.

To generate a base image, simply Start Generating using your style model. For the most consistent results, opt for 1:1 aspect ratio images at this stage. You'll have an opportunity to expand the image size later in the workflow.

When generating your base, it's recommended to use simple prompts, such as "a scene with two characters talking" or "a man standing near a woman." As the scene's complexity increases, it can be beneficial to choose an external reference image for better results.

‍

Step 3: Generating your characters

Once you are happy with an image, select Use as Reference - this will upload your image directly to your inference window.

You will then select the Sketch tool and adjust the color using the eyedropper tool to match the background of the image. It’s time to block out one of the character - we’ll come back to them later. Secondly, remove the background of the image.

Finally, we are going to make a little magic with this reference image using a multi image workflow. We will switch to one of the Composed Character Models and impose them onto the visible character in your reference image. We will do this by switching the Reference Image mode to Image to Image and IP Adapter.

Upload a portrait of your character into the IP Adapter setting and adjust the influence of the IP Adapter to roughly 15. Add a prompt describing the character in just a few words, and generate - this will replace the general character in the image with your custom character.

This step can take a few iterations and adjustments depending on the style you are using. Go ahead and click the link below for the full walkthrough of this step.

‍

Step 4: Assembling with Canvas.

The final step is to assemble and clean up your image. By now, you should have all your character images prepped and ready for upload, along with your original reference base. Begin by uploading your reference base to the Canvas by starting a New Project. Choose the composed model associated with the first character you want to integrate.

Next, switch to Upload and import your first character. Align the character with the original reference base and Erase any artifacts that might have been missed by the background removal tool. During this step, mask and regenerate any small details that you might want to fix.

Once you've completed that, upload your additional characters one by one, starting from the furthest in the background of the image and moving toward the foreground. You can easily adjust their order in the Layers side column. Once your characters are in their correct spots, Merge all the layers.

Finally, mask out any areas that look strange in relation to one another to clean up any loose edges. Switch to the general style model you chose for this workflow. This Canvas has some nuances, so for a more in-depth look, click the link for a deep dive.

Conclusion Section

By now you have your first fully customized, multi character image, supported with AI. There are a few tools and workflows to check out that will bring an extra level of refinement to your image. Consider Upscaling your final image to add some extra details and clean up any areas that look out of sync.

In conclusion, by following this comprehensive workflow, you've unlocked the potential to create captivating multi-character scenes that bring your creative vision to life. With your custom models, characters, and styles, you can now generate a rich and diverse array of images that tell compelling stories. As you continue to refine your skills and explore new possibilities, remember that this workflow serves as a solid foundation for your artistic journey.