2 min to read
Pretty Draboat!! - AI Generated Character
Exploring AI-Powered Character Design with Midjourney
Pretty Draboat!! - AI Generated Character
In this post we will be introducing how to use the AI drawing software Midjourney to create character design prototype illustrations for gacha games. First, here are example of the final generated results: The advantage of the Midjourney model is that, compared to Stable Diffusion, its art style is more stable and controllable. Generally, we determine the character’s hair color, eye color, and personality through keywords. At the same time, it’s necessary to finalize the corresponding decorations and the inspiration sources for the design concept.
For example, the prompt for the second image here is:
long blonde hair, straight hair, green eyes, emerald eyes, flower decoration, no background, full body, shoes
Each prompt starts by specifying the hair and eye color. To ensure the hair and eye colors align more closely with our desired result, we can use nouns to represent colors, such as ‘green eyes’ and ‘emerald eyes,’ or ‘blue eyes’ and ‘sky colored eyes’. Then, we can add some representative words for decorations; finally, we specify the command to generate a full-body image. Note the word ‘shoes’ here; generally, commands like ‘shoes’ are needed for the AI tool to generate full-body photos.
Based on the character design images we obtained, we can first create concept designs. It’s important to note that the original images are typically used as input to maintain consistency in the concept design. The concept design doesn’t have to strictly adhere to the character design; we can experiment extensively to find out which atmospheres suit our character the best.
After generating the character concept images, we then select appropriate keyword prompts to generate backgrounds for the character images. It’s important here to use the character’s image as a prompt to maintain consistent color tones:
This is the character we obtained:
Use this photo as prompt, we then add another description like “Chinese Style”, “Painting”. The resulting image should have same color pitch as the character:
Alternatively, we can directly describe the relationship between the character and the background. This approach can also yield images of our character within various backgrounds:
We use this image and the following word prompt:
Ancient Chinese style, wooden bar, windows, candles, indoor, room, walls
Yields the following result: