Step-by-Step Guide: Consistent Characters & Screens with Midjourney + Nano Banana

This week I played with keeping things consistent: same character from different angles, same spot but with different people. Here’s the workflow I used with prompts and examples.

Step 1. Build a Character & Vibe in Midjourney

Midjourney is still the best platform I’ve found for creating strong source images that later expand into a story or concept.

For this photorealistic concept, I used a prompt like:

fashion shot of a modern woman [ADD DETAILS], casual, minimalist --chaos 100 --ar 9:16 --exp 100 --sref 2216056865 4129353691 1742826451 3368914030 689668977 1639812559 2825562739 --profile de2v2uv gy97k8r gtbhewc gj5325t

Note:

• --chaos 100 and --exp 100 are very high values. They give you unique characters/outfits, but you’ll need many reruns.

• I usually run permutation prompts with --r 40 (40 prompts at once). The number of prompts you can run simultaneously depends on your plan.

• If you want faster, more predictable results, lower these parameters (e.g. --c 20 --exp 30).

From the grid, I picked my main character.

Step 2. Generate Different Angles of the Character

Once the character is set, I need to explore different angles and outfits.
You can try this in Midjourney, but right now Nano Banana (Google’s model) does it better.

I use it through Krea.ai:

• Pick Nano Banana (bottom left corner).

• Upload your Midjourney image.

• Add a simple prompt like: ‘woman, full body shot‘

Results:

Nice! I like the second one. This step gives a full outfit view, making it easier to generate consistent angles and shots later.

Step 3. Refine With Variations

With the full-body base image, I remove the old one from the prompt bar, add the new one, and start prompting details.

Example prompts:

• ‘woman, different angle, macro, hands‘

• ‘woman, different angle, macro, shoes‘

• ‘woman, different angle, macro, earrings‘

Step 4. Expand the Scene

Using the same reference, I start adding objects—like a chair.
I’ll reuse this chair in later steps and with other characters.
This keeps everything looking like one consistent world.

You can do the same with other interior elements so they match the character’s mood and outfit.

Prompt: ‘a chair in an empty room in the style of the image.’

Step 5. Add a Character to the new Scene

Take the previous image and the new one with the chair, combine them, and add the prompt: ‘a woman sits on a chair.’

Step 6. Build a Shared World

Repeat the process to generate multiple characters in the same scene.
They’ll naturally feel like they belong to one coherent “world.”

---

That’s the workflow! Once I had a big library of assets, I started using them to make videos.
I’ll dive into that in the next article.

Happy prompting ✨