Master Stable Midjourney Characters: 5 Commands for Consistent AI Art

Introduction

Previous articles showed how to create a quick comic in 15 minutes, but Midjourney’s randomness often yields different results for the same prompt. To generate stable characters with specific expressions, actions, and clothing, five reliable commands are presented.

Base Prompt

Blind Box Style Girl Delicate features, With a cat ears hat, furry clothes, full body, looking at viewer, Super Details, edgelighting, chiaroscuro, 8K, chibi, best quality

Part 1: URL + Initial Command

Create a favorite character in Midjourney; the bot generates an image (Fig. 3).

Upscale and save the desired image to your computer (Fig. 4).

Upload the saved image and click the plus (+) button in the chat box, then press Enter.

Open the image, copy its URL.

Enter the command /imagine prompt +URL+initial command to generate a new image that retains the original style.

Part 2: Approximate Description + Seed Command

To produce a series of images from a single photo, obtain a seed value:

Generate a satisfactory image.

In the four‑image grid, click the reaction (emoji) button.

Search for and click the envelope icon.

The Midjourney bot sends a message containing the seed value.

Modify the original command by appending --seed <seed_number> and resend.

Part 3: Panels Command

Use the panels command to create consecutive poses: 4 panels with different poses + base prompt → generates four images of the same character in different actions (Fig. 7).

For continuous motion, use 4 panels with continuous eating + base prompt to obtain a sequence of the character eating.

Part 4: Character Sheet Command

Generate a full character sheet with various angles and accessories: character sheet + base prompt → produces a set of images showing the character from multiple perspectives (Fig. 9).

Part 5: Emoji + Expression Sheet Command

Use the emoji+expression sheet command to obtain a range of facial expressions for the same character: emoji+expression sheet + base prompt → creates multiple expression images (Fig. 11).