Text to Image

Overview

The Text to Image tool allows you to type a prompt, and the model will generate images using your trained character/IP in your animation style.

Because the model already understands your character, you do not need to describe your IP again. Simply refer to the character by "name".

What Makes a Good Prompt?

A strong prompt includes:

A clear action or scenario

• “Running through a forest”

• “Reading a book at a cozy café”

Emotional tone or expression

• “Excited and smiling”

• “Surprised, arms raised”

Environmental or stylistic details

• “Soft autumn lighting”

• “Dim evening lighting”

You do not need:

• Face/body attributes

• Character details

• IP descriptions

Atrium already knows your character after training.

You can then choose the output orientation: portrait, landscape, or square. Here you will also pick how many generations you’d like it to make (1 through 4).