1. What Makes v6 Different
Midjourney v6 represents a structural overhaul in natural language processing. Unlike previous versions (which required comma-separated keywords like "8k, photorealistic, rendering engine"), v6 is optimized for plain English prose. It reads sentences structurally, understanding the relationship between verbs, adjectives, and nouns.
If you clutter a v6 prompt with outdated hype words, the engine will often get confused or ignore the core artistic directives.
---
2. The Power of Style Raw
By default, Midjourney applies its own artistic signature to your prompts. While this is often visually pleasing, it can override your specific creative directions.
By appending `--style raw` to your prompts, you bypass Midjourney's built-in styling, giving you a literal, high-fidelity interpretation of your text.
#### Comparison:
Default:* "A vintage portrait of an astronaut." -> Generates a highly stylized, cinematic, slightly idealized illustration.
With `--style raw`:* "A vintage portrait of an astronaut. --style raw" -> Delivers a more accurate, documentary-style photograph matching authentic historical emulsions.
---
3. Fine-tuning with --stylize
The `--stylize` (or `--s`) parameter controls the strength of Midjourney's artistic training. The range goes from 0 to 1000:
---
4. Absolute Character Consistency (--cref)
One of the most powerful features in v6 is Character Reference (`--cref`). This allows you to generate the same character across multiple completely different scenes.
#### How to use:
1. Generate or upload your base character image.
2. Copy the direct URL of that image.
3. Write your new scene prompt, and append `--cref [IMAGE_URL]`.
4. Adjust character weight with `--cw` (from 0 to 100). `--cw 100` keeps face, hair, and clothing consistent, while `--cw 0` only focuses on the facial structure.