🧠How I keep the character consistent across multiple AI images

Omni Reference in V7 for Consistent Character Design

May 09, 2025

Hello there! 👋 
If you've ever used AI tools to generate images, you know how tricky it can be to get consistent results. One moment your character looks perfect, and the next they look completely different.

This week, we’re diving into Midjourney Version 7’s Omni Reference, a new feature that helps you keep the same character or object across multiple scenes without losing their core look.

Whether you're working on a comic, building brand visuals, or just experimenting with AI art, this guide covers:
✅ Using Omni settings in Discord and the Web UI
✅ Adjusting how closely your image follows the reference
✅ Adding style with moodboards, effects, or your own touch

No design background needed. Just an image, a prompt, and a little curiosity.

Let’s get started.👇

🧩 The Tech Stack: What’s What

Web Image Upload

✅ All-in-One Midjourney Setup Workflow

 🧠 1. Pick the Right Reference Image

  • Clarity: The character or object should be clearly visible, centred and well-lit. 

  • Image Quality: Use high-resolution, sharp images.

  • Background: A plain or transparent background works best.

  • Multiple Characters: You can use more than one, just make sure you describe each in your prompt.

🧠 2.Set it Up in the Web UI

  • Choose Model V7.

  • Upload your reference image and Drag into the Omni Reference section.

  • Write your prompt (describe the scene, mood, and style).

  • Use the omni weight slider or type --ow  followed by a number. 

  • Click Generate.

Make sure “Omni-Reference” appears in the top-right panel

3. Set Up in Discord

  • Type /settings and choose V7.

  • Upload image → Right-click → "Copy image address."

  • Add to prompt like:

    “A futuristic detective in a neon-lit city --oref https://image-url.jpg --ow 400”

  • Send the prompt.

    3. Set Up in Discord

Weight Range

Effect(s)

1–50

Loose reference; more artistic freedom

50–150

Balanced look and flexibility (Default is 100)

200–400

Strong resemblance to your image

500–1000

Near-exact likeness, great for logos or portraits

This slider controls how strongly the Omni-Reference affects your image

🧠5. Bonus Tips for Customization

✅ Add a style: Use --style for watercolor, anime, or realism. You can also upload a second image to guide the art direction.

✅ Use a moodboard: Upload 2–3 images that show your desired aesthetic (color, lighting, etc.). Midjourney blends them into the result.

✅ Personalize it: You can use your own photo, your pet, or your brand elements to give the generation a unique touch.

✅ Make it cinematic: Want a film-like finish? Add --exp 5 to your prompt for bold lighting, enhanced contrast, or dreamy detail.

🎭️ Omni Reference: Pros & Limitations

Pros

Limitations

Maintains character or object consistency across multiple generations

Only works with Version 7

Supports personal styles, moodboards, and reference-driven customization

Not compatible with Pan, Zoom, Vary Region, or Inpainting

Works in both Discord and Web UI

Uses 2× GPU time, leading to higher cost and longer rendering

Gives you control over how closely the output matches your reference

Allows only one reference image per prompt

Ideal for comics, branding, storytelling, and concept design

Results depend heavily on image quality — blurry inputs may affect output

AI Challenge of the Week! 🐅

Let’s put Omni Reference to the test, with a Met Gala twist ✨

👗 Step 1: Upload an image of your favorite celebrity or fictional character.
🎨 Step 2: Style them like they’re walking the Met Gala red carpet.
🔥 Step 3: Use this prompt (replace the image URL with your own):

Try this prompt: 

[Your character] walking the Met Gala red carpet in an avant-garde futuristic outfit, spotlight, flash photography, highly stylized fashion editorial --oref https://your-image-url.jpg --ow 600 --style raw --v 7


Challenge us back! 💪
Reply to this email with your best GenAI Met Gala look or upload it on social media and tag @DigitalSamaritan

💌 The most jaw-dropping look will be featured in our next newsletter!

👀 ICYMI

Think you need to code to build an app? These AI tools prove you don’t - Watch how anyone can create powerful apps with zero experience. 👉 See for yourself!

What happens when AI and humans go head-to-head in creating user content? The results of this experiment might surprise you! 👉 Check Here! 

AI Roundup

  • This new AI from Hugging Face uses a real computer to complete tasks — but can it survive Google Maps and CAPTCHAs? 👉 Dive in!

  • This AI doesn’t just talk about medicine. It sees it, diagnoses it, and often performs better than human doctors in tests. 👉 Read the full story 

Did you learn something new?

Login or Subscribe to participate in polls.

💌  We’d Love Your Feedback

Got 30 seconds? We'd love to know what sparked your curiosity, or what didn’t. Hit reply and spill the beans!

Until next time,
Team DigitalSamaritan

Reply

or to participate.