Composing and posing a character

Introduction

In this article, we will delve deeper into creating a new character using one of our platform models and posing the character using a reference image. This workflow serves as a building block for various use cases, ranging from dress-up games to sprite sheets. It offers an alternative to training a character model and is an excellent option for developers who do not have a specific character in mind.

Overview

  1. Compose your base models
  2. Generate portraits
  3. Generate pose using dual reference

Compose your base models

To create a new character using one of our platform models, first navigate to Models on the sidebar and press the + button. Then, click Start Composing. Select one or more characters you want to compose by clicking Tags and choosing Character Base. After making your selection, clear out the tag and choose a style to add.

To test the settings of your model and achieve the desired style, click Test and use a simple prompt. Once you're satisfied with your composition, close the test window. You can also add and remove model components as needed. Finally, name your model and click Save Composition.

Generate a portrait

To begin generating images with your new model, first locate it in the Models section under Your Models. Click on Generate with This Model to proceed.

In the prompt, type "portrait of" and add any necessary descriptions, such as "a portrait of a woman smiling." After generating images, click on the one you like the most. To iterate further, click Use as Reference, or move on to the next step in your workflow.

Generate pose using dual reference

To achieve the desired pose for your character, first find a reference image that showcases the pose you'd like to use. Describe the pose in the prompt, using terms such as "T Pose" or "A Pose." Then, upload the reference image into the Reference Image section.

Next, switch to Controlnet + IP Adapter and change the Mode of Controlnet to Pose. Although "character" is an alternative that can work, it may alter some physical details of your character, such as face shape. Click the "…" symbol and select Remove Background, then set the pose influence to 100.

Upload a portrait of your character to the IP Adapter and adjust its influence to 10. Add any additional descriptors of your character to the text prompt, such as clothing details. Generate the image and adjust the settings as needed. If the face is not detailed enough, slightly increase the IP Adapter influence.

(Optional) Refining Details with Upscale

To enhance face details in your generated image, consider applying a Balanced upscale. Select an image you have generated, click Upscale, and choose a style. Generate your upscaled image, and if the style is not consistent enough with your original, opt for the Precise option. Alternatively, if you want more details added, select the Creative option.

Table of content

You'll also like