One of the hardest challenges in AI image generation has always been keeping a character looking the same across multiple scenes. A hero in a fantasy landscape, then at a coffee shop, then in a business meeting — most AI models treat each prompt as a blank slate. Midjourney V7 changes this with two powerful tools: Omni Reference and the classic –cref parameter. Here is how to use them in practice.
What Changed in V7 for Character Work
Midjourney V7 launched on April 3, 2025 and became the default model on June 17, 2025. Compared to V6, V7 delivers a 40% reduction in anatomical errors — hands, faces, and body proportions are noticeably more reliable. But the real breakthrough for character artists and storytellers is how V7 handles named character references across a session.
According to Midjourney official documentation and confirmed by the V7 release announcement, V7 introduces Omni Reference as a first-class feature: you name a character, provide a description once, and V7 consistently applies that character across all subsequent images in the session. This makes visual storytelling —comics, sequential art, brand mascot consistency — far more practical than before.
Method 1: Omni Reference (Session-Level Consistency)
Omni Reference is the newer, more powerful approach introduced with V7. It works at the session level, meaning you set it up once and it applies to every image you generate in that session.
Step 1: In your Midjourney prompt box, type your character description and name it. For example: A woman in her 30s, sharp cheekbones, shoulder-length dark hair, warm brown eyes, confident posture — named CHARACTER_NAME
Step 2: Generate your first image with this prompt. V7 will register this character definition for the session.
Step 3: In subsequent prompts, simply refer to the character by your chosen name: CHARACTER_NAME sitting in a Tokyo café, reading a book
V7 will apply the same facial structure, hair, and general appearance across scenes. This works across different settings, lighting conditions, and outfits — as long as you keep the character name consistent.
Method 2: Character Reference Flag –cref (Per-Image Consistency)
The –cref parameter has been available since earlier Midjourney versions and remains useful in V7. Unlike Omni Reference, –cref targets a specific reference image you provide, making it useful when you have an existing character design you want to preserve.
Step 1: Upload your reference character image to Discord or use an image URL.
Step 2: Add --cref [image_url] to your prompt. You can also specify intensity with --cw 0-100 (default is 100, meaning full face and body; lower values apply mainly to face).
Step 3: Example full prompt: A woman in a yellow trench coat standing on a Hong Kong rooftop at dusk --cref [url] --cw 80 --ar 16:9 --v 7
The –cw parameter is useful when you want the character in a very different outfit or setting — a lower value keeps the face consistent while allowing more creative freedom with the rest of the composition.
When to Use Which Method
Use Omni Reference when you are building a story or brand identity and want consistent results across many different scenes without re-uploading a reference image each time.
Use –cref when you have a specific character design already created that you need to replicate precisely, or when you want fine-grained control over how much of the reference influences the output.
Draft Mode: Iterate Faster at Half the Cost
V7 also introduces Draft Mode, which generates previews at 10x the speed and half the cost of standard rendering. Before committing to a final upscale on a character scene, switch to Draft Mode to quickly check if the character consistency is working as intended. This dramatically speeds up the iteration cycle when you are fine-tuning a character across multiple scenes.
What Still Has Limits
Even with V7 improvements, very dramatic angle changes and extreme lighting shifts can still cause subtle variations in how the character reads. Extremely long sessions with many generations may occasionally drift. For print-quality commercial work, it is still wise to review each output carefully. The –cw parameter gives you control — full body consistency is harder than face-only consistency.
Putting It Together
Start a new Midjourney session. Define your character with Omni Reference. Generate your first scene. Then iterate through your story beats — coffee shop, office, street — and use Draft Mode for quick checks. When you have a scene you want to keep, upscale with the full –v 7 render. The combination of session-level character memory and per-image fine-tuning gives you more control over visual consistency than any previous Midjourney version.
