After about a year, I returned to Midjourney to test its capabilities for AI comic creation. Unfortunately, I have to say that I’m disappointed by the lack of progress.

I used Version 7 for this project. While I was working on the comic, Midjourney released V8, but I was already about 90% finished with V7, so I decided to stick with it. From what I’ve seen so far, V8 doesn’t appear to offer significant improvements in the areas that matter most for comic creation: character consistency and environment consistency.

If you’d like to check out the comic, you can find it here: The Last Superhero Part 5

How to Create AI Comics with Midjourney

Create Master Prompts

You want your style, characters, and settings to be defined through master prompts. Keep them as short as possible to reduce prompt fading.

Examples:

• Protagonist: mature man, grey hair, grey beard, stern face, dark coat, layered dark clothing, tall
• Female lead: woman in her 30s, dark hair, slim build, worn clothing, subtle manipulative expression
• Style: comic, Scott Snyder style, black and red colors
• Mountain setting: mountains, canyon road, military convoy driving

Go through your script and create master prompts for all characters and settings involved. The style master prompt should be added to every image prompt.

This approach doesn’t guarantee consistency, but it improves it as much as currently possible in Midjourney.

Create Character Image Sheets

Next, create character sheets for your main characters. For my comic, I created sheets for the protagonist, the female lead, and the NPC soldiers. At a minimum, I recommend creating the following:

Examples:

• Character Main Sheet: [style master prompt + character master prompt], full-body shot
• Character Face Sheet: [style master prompt + character master prompt], multiple panels, different facial expressions (or angles), plus specific emotions if needed (e.g., surprised)

The main sheet serves as your primary character reference. You can upload these images to Midjourney and use them as character references whenever that character appears in a scene.

The face sheet can be used as an overlay. You can place it on top of larger scene images whenever a character is speaking. I’ve used this technique many times.

Use Character Sheets in Two Ways

Once you’ve established a consistent appearance for your characters, crop the image in two different ways:

1. Full-body shot: Remove as much of the background and non-character details as possible.
2. Face shot: Crop tightly around the character’s face.

You can then use both images as character references, which helps improve consistency.

Tips from 2025 Still Apply

If you haven’t read my guide from last year, nearly all of the advice still applies: How to Make Comics with AI (Midjourney | 2025)

Problems with Creating AI Comics with Midjourney

My previous attempt using ChatGPT produced significantly better results. The main reason is that Midjourney still hasn’t solved the two most important issues:

1. Consistency
2. Generating multiple characters in the same image

Character consistency remains a major problem, even with the option to upload reference images. Style is difficult to control and can easily fall apart, especially when prompts become longer. There is also still no reliable way to maintain environmental consistency across scenes.

Even worse, there has been virtually no improvement when generating images with multiple characters. It seems that using several character references causes features to blend together. The same issue occurs with character actions. As a result, it’s extremely difficult to create scenes where two characters interact naturally while maintaining consistent appearances from panel to panel.

Other Problems

Language Filter
You still can’t use many words associated with violence. Since most action scenes involve some form of violence, this limitation makes it difficult to tell traditional comic-book stories.

Prompt Fading
The longer your prompts become, the more tokens tend to “fade.” In practice, this means Midjourney starts ignoring certain parts of your prompt.

Style Drift
Maintaining a consistent visual style was even more difficult than a year ago. If you compare the first pages with the final pages of TLS 5, you’ll notice a significant style drift that I wasn’t able to control.

Improvements Compared to Earlier Midjourney Versions

Faster Image Generation

Midjourney has become even faster. You can generate roughly 500 images per hour, giving you plenty of options to choose from. This remains a significant advantage over ChatGPT’s image generation.

High-Quality Individual Images

When focusing on individual images, the quality is excellent. With enough patience, it’s possible to create impressive standalone comic pages.

Conclusion

Overall, there hasn’t been much progress—especially in the areas that matter most. Consistency remains a major challenge, and the language filter makes action scenes unnecessarily difficult to create.

In many respects, ChatGPT’s V5 image generation already produced better results for comic creation. My next step will be testing ChatGPT’s newest image model, which I’ve heard very positive things about. Perhaps that’s the model that finally brings us a significant step closer to creating truly convincing comics with AI.

I went back to using Midjourney with this part. Since my first try with it in 2025 a couple of new versions dropped. I’ve mainly used V7 for the following comic. A full review of Midjourney V7 will be published soon.

The Last Superhero 5