How to Architect Consistent Brand Visuals with Midjourney v6 (The -cref Protocol)

⚡ Workflow Specs

Tools Required: Midjourney v6 (Discord or Web Alpha)
Time: 15 Minutes
Cost: Paid (Standard Plan recommended for GPU hours)
Difficulty: Advanced

Contents hide

1 The Scenario

2 Step-by-Step Instructions

2.1 Phase 1:

2.2 Phase 2:

2.3 Phase 3:

2.4 Phase 4: Style Transfer (Optional)

2.5 ⚠️ Expert Note on Troubleshooting:

3 Stay Ahead of the Curve

4 Related

5 Discover more from Technosys Blogs

The Scenario

Standard AI image generation creates “random” faces, rendering it useless for sustained brand storytelling or storyboards. You don’t need a person; you need the person, your specific brand persona, appearing consistently across 50 different assets, angles, and lighting conditions without morphing into a different human.

Step-by-Step Instructions

Phase 1:

Generating the “Anchor Asset”

Technical Insight: The “Anchor” is the source of truth for the neural network. We need a neutral, evenly lit shot to serve as the reference map. Do not use artistic angles here.

Launch your private Midjourney Direct Message or server.

Execute this specific prompt to generate your character (replace bracketed text):

/imagine prompt: [Subject Description, e.g., A 35-year-old software engineer, Scandinavian features, glasses, sharp jawline], studio lighting, neutral grey background, shot on Sony A7R IV, 85mm lens, hyper-detailed texture --ar 3:4 --style raw --v 6.0

Upscale (U1/U2/U3/U4) the best result.
Right-click the upscaled image and select Copy Link (ensure the link ends in .png or .jpg).
Save this URL in a notepad. This is your [ANCHOR_URL].

Phase 2:

The Character Reference (--cref) Injection

Technical Insight: We will now force the model to reference the latent features of your Anchor URL using the --cref parameter. We will also control the “Character Weight” (--cw) to dictate how much influence the reference has.

Type /imagine but do not hit enter yet.
Construct your prompt using the following syntax structure. This specific order is critical for token parsing:
```
[New Scene Description] --cref [ANCHOR_URL] --cw [WEIGHT] --v 6.0
```
Determine your --cw (Character Weight) value:
- Use --cw 100: To lock everything (Face, Hair, and Outfit). Best for same-day continuity.
- Use --cw 0: To lock only the Face (Allows you to change outfits/hair). Best for different campaign scenarios.

Phase 3:

Batch Production via Permutations

Technical Insight: Instead of generating one by one, we will use Permutation Prompts { } to run concurrent jobs. This is how pros generate full libraries in minutes.

Copy and Paste this code block to generate the character in 3 distinct business scenarios simultaneously (this consumes fast hours):

/imagine prompt: A photo of the character {working on a laptop in a modern office, presenting on a stage with a microphone, shaking hands in a boardroom meeting} --cref [PASTE_ANCHOR_URL_HERE] --cw 0 --ar 16:9 --style raw --v 6.0

Hit Enter and Confirm the batch job.

Phase 4: Style Transfer (Optional)

11. If the lighting doesn’t match your brand palette, find an image that does have the perfect vibe. Copy its link ([STYLE_URL]).

12. Append the --sref (Style Reference) parameter to the end of your prompt chain: text ... --cref [ANCHOR_URL] --cw 0 --sref [STYLE_URL] --sv 2 --v 6.0

⚠️ Expert Note on Troubleshooting:

If the face starts to distort (the “burnout” effect), it usually means your --stylize value is fighting the --cref. Reduce --stylize to --s 50 or remove it entirely to prioritize facial fidelity over artistic flair.