Midjourney v6 is the most significant model update since the platform launched. With dramatically improved text rendering, better nuance, more realistic textures, and two new reference systems (--sref and --cref), it's now capable of outputs that were impossible just a year ago. If you're still using v5 prompting techniques, you're leaving 80% of v6's power on the table.
The New v6 Prompting Philosophy
V6 was trained on a fundamentally different dataset and responds best to conversational, descriptive prompts. The old approach of stacking quality keywords like "8k, ultra-detailed, photorealistic, trending on ArtStation" no longer works — in fact, it can degrade output quality.
Old v5 approach (don't do this):
New v6 approach (do this):
The 5 Most Important v6 Parameters
--sref [image URL](Style Reference): Locks the visual style of a reference image while changing the subject. Invaluable for brand consistency and style series.--cref [image URL](Character Reference): Maintains a specific character's face and appearance across different scenes. Game-changer for illustrated storytelling.--stylize [0-1000]: Controls how much Midjourney's own artistic "opinion" influences the output. Default is 100. Higher values (700+) give more artistic flair; lower values (0–50) stick more strictly to your prompt.--chaos [0-100]: Increases the variety and unexpectedness of the initial image grid. Great for creative exploration when you're not sure what you want.--tile: Generates seamless, repeating patterns. Essential for textile design, wallpapers, and website backgrounds.
Text Rendering: The v6 Superpower
For the first time, Midjourney can accurately render short text strings within images. Wrap your text in quotation marks inside the prompt:
The New Up-scalers
V6 introduced two new up-scaling modes after you generate your initial image:
- Subtle Up-scale: Increases resolution (up to 1024px) while keeping the composition and details exactly as generated. Best for when you love the result.
- Creative Up-scale: Adds significant new detail and reinterprets elements at higher resolution. Best when the initial image is close but needs more refinement.
Pros
- Unmatched aesthetic quality in any AI image tool
- Excellent style and character consistency with --sref/--cref
- Robust text rendering for shorter strings
- Active community with prompt-sharing on Discord
Cons
- Subscription-based with no free tier
- Discord interface has a steep learning curve for beginners
- Cannot generate images with explicit or copyrighted content
- Complex scenes with many characters still struggle with consistency
Pro Workflow: From Idea to Final Image
- Start with
--chaos 50to get varied initial directions. - Pick your favourite variant and run "Vary (Subtle)" to refine.
- Once satisfied, run "Upscale (Subtle)" for maximum resolution.
- Save the final image URL and use it as a
--srefin future prompts to lock the style.
Final Thoughts
Midjourney v6 is more than just an image generator — it's a complete visual ideation engine. By mastering the new conversational prompting style and leveraging the --sref and --cref systems, you can unlock a level of creative consistency that was previously only possible with a dedicated human illustrator.