Midjourney v6.1 Parameter Cheat Sheet

What Do the Core Midjourney v6.1 Parameters Actually Do?

Use parameters after the descriptive prompt. The table prioritizes the controls most often needed for predictable commercial, editorial, and exploration runs. It follows the official Midjourney parameter list as its source of truth, then narrows each control to a production starting point rather than documenting every accepted value.

Advanced Style & Character Consistency: --sref and --cref

--sref tells Midjourney which visual language to echo: palette, material, lighting, and image-making feel. --cref is the identity anchor for a recurring subject. They solve different problems, so use both only when the campaign needs a stable person and a stable art direction. Midjourney's documentation covers the style reference and character reference behaviors in depth; the recipes below compress that into copy-ready syntax.

Need a strong reference image first? Run it through Magic Reverse image-to-prompt, then save the resulting prompt and reference URL together in the Midjourney Prompt Library.

Tuning the weight values

--sw (style weight) and --cw (character weight) are the two dials people most often over-crank. Start low and raise only the control that is visibly drifting: if the palette is wandering across a campaign, nudge --sw up 20–40 points at a time; if a character's identity holds but the wardrobe is frozen, lower --cw toward 30–50 rather than pushing it to 100. The guiding rule is that these weights compete with your descriptive prompt — the higher they go, the more the reference overrides what you wrote. When a brief must remain literal, keep the weight at its floor and put the detail back into the text prompt itself.

A repeatable iteration loop

Consistency is a process, not a parameter. A disciplined loop looks like this: fix the model (--v 6.1) and the destination ratio (--ar) first, since those two changes cascade into every other decision. Then change one control per grid — stylize, chaos, or a single reference weight — and save the winner with its full syntax, seed, and reference URL before touching anything else. Compare grids side by side, delete the noise, and only bump --q or --repeat once the direction is locked. This turns a parameter reference from a list of knobs into a workflow you can hand a colleague and trust they will reproduce your best results.

What Are the Optimal Midjourney v6.1 Parameter Combinations?

Production Asset Case Studies: E-commerce & Concept Art

Parameter tables teach you the controls; case studies show you how they hold together under a delivery deadline. The two scenarios below — a direct-to-consumer product set and a game-asset concept sprint — use the same matrix but weight the controls very differently, which is exactly the point: there is no universal "best" value, only the value that serves the placement.

The pivot between the two

The e-commerce run pushes consistency parameters (--s, --c, --no) toward their floor; the concept run pushes variation parameters (--c, --weird) toward their ceiling. Everything else — --v 6.1 for a stable model baseline, --style raw for honest materials, and a recorded --seed or --sref for reproducibility — stays fixed. That discipline is what separates a repeatable production workflow from a lucky image. The parameter documentation remains the authoritative reference whenever a value changes by model or mode.

Pitfalls that break production runs

1. Bumping quality before locking the composition. Fix: iterate at --q 1 and reserve --q 2 for the final approved frame; higher quality costs more without improving an unproven layout.
2. Mixing a brand reference with a high character weight. Fix: a product set needs a stable --sref, not a --cref pulling in a recurring figure; keep --cref for character-led storyboards and lower --cw when wardrobe should adapt.
3. Batching an untested prompt with --repeat. Fix: confirm a single grid first, then multiply with --repeat; each queued run bills against the same idea, so a bad prompt becomes many bad outputs.

Which Syntax Mistakes Break Midjourney Prompts — and How Do I Fix Them?

1. Putting parameters inside the description. Fix: write descriptive language first and place --ar 4:5 --s 180 at the end, separated by spaces.
2. Changing aspect ratio after approving a composition. Fix: select --ar from the destination format before comparing iterations; a new ratio asks for a new composition.
3. Maxing stylize, chaos, and weird together. Fix: assign one parameter to the experiment. For example, hold --s 200 and --weird 0 while testing --c.
4. Using a vague or inaccessible reference URL. Fix: use a direct image URL that Midjourney can access; record which source image powered each approved result.
5. Expecting --cref to preserve every detail. Fix: describe pose, clothing, framing, and mood in the text prompt, then use --cw only to tune how tightly identity is held.

Frequently Asked Questions