Midjourney
Midjourney V6 Parameters Cheat Sheet: --style raw, --q, --v
Every Midjourney V6 and V6.1 parameter in one table: --v 6.1, --style raw, --q 2, --s, --c, --ar, --sref, upscalers, plus copy-paste prompts.
By Eren Koray. Checked against the sources listed at the end on . 7 min.
On this page
Midjourney V6 and V6.1 are older models, but you can still choose them for any prompt. This page lists every parameter that works with them, the accepted values, and ready-to-copy prompts.
Current default: V8.2 (since 24 July 2026). If you do not add --v, your prompt runs on V8.2 or on whatever version you picked as your default in settings. To use V6.1, add --v 6.1 to the end of the prompt.
V6 / V6.1 parameter cheat sheet
| Parameter | Short form | Values in V6 | Default | What it does |
|---|---|---|---|---|
| Version | --v 6 / --v 6.1 | 6, 6.1 | V8.2 | Chooses the model for this prompt |
| Raw | --style raw or --raw | on / off | off | Turns off Midjourney's automatic styling for a more literal, photo-like result |
| Quality | --q | 0.5, 1, 2 | — | More GPU time on the first grid for more detail |
| Stylize | --s | 0–1000 | 100 | How strongly Midjourney's own aesthetic is applied |
| Chaos | --c | 0–100 | 0 | How different the four images in a grid are from each other |
| Weird | --w | 0–3000 | 0 | Unusual, offbeat aesthetics (experimental) |
| Aspect ratio | --ar | e.g. 16:9, 2:3, up to 14:1 | 1:1 | Image shape |
| No | --no | words | — | Removes elements; same as weighting them ::-0.5 |
| Seed | --seed | whole number | random | Repeats the starting noise for closer re-runs |
| Tile | --tile | on / off | off | Seamless, repeating pattern |
| Image weight | --iw | 0–3 | 1 | How much an image prompt counts against the text |
| Style reference | --sref | image URL or code | — | Copies the style of a reference image |
| Personalization | --p | profile code | — | Applies your personalization profile (V6 and later) |
| Niji 6 | --niji 6 | — | — | The anime model of the V6 generation |
The values come from Midjourney's Version comparison chart and Legacy Features table (links at the end). Quality 0.5, 1, 2 is the V6 set: V7 uses 1, 2, 4, and the Version chart does not list Quality for V8.1/V8.2.
How to write the parameters
Midjourney reads everything after the first -- as parameters. The rules from the official Parameter List:
- Write the description first, parameters at the end.
- Put a space before every
--. - No commas or other punctuation inside parameters.
- Do not continue the description after the parameters start.
a lighthouse on a rocky coast at dusk, long exposure, soft fog --ar 16:9 --style raw --q 2 --v 6.1
Common mistakes:
a lighthouse at dusk--v 6.1 (no space before --)
a lighthouse --v 6.1, soft fog (text after the parameters)
a lighthouse at dusk --ar 16:9, --q 2 (comma between parameters)
a lighthouse at dusk —v 6.1 (an em dash instead of two hyphens)
The last one happens on phones and Macs that turn -- into a long dash automatically. If Midjourney ignores a parameter, check this first.
--style raw vs --raw
Both forms appear in Midjourney's documentation. The Legacy Features table lists Style with the value raw for V5 and V6, so most V6-era prompts use --style raw. The current Raw page documents --raw and says it works on Midjourney 5.1 and later. You can also switch Raw on as a default in the settings panel.
Use Raw when Midjourney adds drama you did not ask for, or when you want a plain photographic look:
candid photo of an elderly baker dusting flour off a counter, morning window light, 35mm --style raw --v 6.1
product shot of a matte black water bottle on white seamless background, soft box lighting --style raw --s 50 --v 6.1
Raw with a low Stylize (--s 50 or lower) gives the most literal reading of the prompt.
--q 2 in V6
Quality changes how much GPU time goes into the first grid. In V6 the accepted values are 0.5, 1 and 2.
--q 2costs more GPU time and can add detail.--q 0.5is faster and cheaper, which is handy for testing compositions.- Quality only affects the first four images. It does not change variations, Vary Region, Zoom Out or upscales.
macro photo of frost on a red maple leaf, fine ice crystals, overcast light --q 2 --v 6.1
rough composition test, knight on horseback crossing a river --q 0.5 --v 6.1
Stylize, Chaos and Weird
These three are easy to confuse:
| You want... | Use | Try |
|---|---|---|
| Less "Midjourney look", closer to the words | Lower Stylize | --s 0 to --s 50 |
| More artistic, painterly polish | Higher Stylize | --s 250 to --s 750 |
| Four very different ideas in one grid | Chaos | --c 30 to --c 60 |
| Strange, unconventional images | Weird | --w 250 to --w 1000 |
The settings presets map Stylize Low/Med/High/Very High to 50/100/250/750, which makes those values a good starting ladder.
watercolor map of an imaginary island, hand-lettered labels, sea monsters in the margins --s 750 --v 6.1
a futuristic tea house --c 50 --ar 3:2 --v 6.1
portrait of a fox wearing a knitted scarf, studio lighting --w 500 --v 6.1
Upscaling a V6 image
V6 has no upscale parameter. You upscale after generating, with two buttons:
- Upscale (Subtle): doubles the size and keeps the image as close to the original as possible.
- Upscale (Creative): doubles the size and adds detail. It can fix small problems such as an awkward hand, and every re-run is slightly different.
Each upscale uses GPU time. Midjourney's Upscalers page notes that V7 images also use the V6.1 upscalers, and the Zoom Out page says zooming still runs on V6.1 for V7 and V8.1 images.
Seed, Tile and --no
Repeat a result as closely as possible by reusing its seed:
isometric cozy bedroom, pastel colors, soft shadows --seed 12345 --v 6.1
Make a seamless pattern for fabric or a website background:
hand-drawn lemons and leaves, flat illustration, cream background --tile --v 6.1
Remove something you keep getting:
empty mountain road at sunrise, wide landscape --no cars, people, text --ar 16:9 --v 6.1
--no cars works like giving "cars" a weight of -0.5. For more on weights such as cat::2, see the prompt weighting guide.
Style references and the old Character Reference
--srefworks with V6 and later. To use old style reference codes, Midjourney's Style Reference page says to add--sv 4or switch to V6.--cref(Character Reference) is now a legacy feature. It was used with V6 and Niji 6. Newer versions replace it with Omni Reference (--oref), which does not work on V6.
a girl reading in a train compartment, golden hour --sref https://example.com/style.png --v 6.1
Replace the URL with a link to your own reference image.
What does not work in V6
The Version chart marks these as unavailable for V6. Remove them when you add --v 6.1 to a prompt written for a newer version:
| Parameter | Available from |
|---|---|
--oref, --ow (Omni Reference) | V7 |
--draft (Draft mode) | V7 |
--hd | V8.1 |
--exp | V7 |
V6 vs V6.1 at a glance
| V6 | V6.1 | |
|---|---|---|
| Released | December 2023 | 30 July 2024 |
| Default until | — | 16 June 2025 |
| Notes | More accurate with longer prompts | More coherent images, finer detail and texture, about 25% faster than V6 |
| Select with | --v 6 | --v 6.1 |
Build your prompt
The prompt generator builds the descriptive part of a prompt from your subject, style and lighting. Paste it into Midjourney and add the parameters from the table above. It runs in your browser and does not generate images.
Primary sources
- Midjourney: Version
- Midjourney: Parameter List
- Midjourney: Legacy Features
- Midjourney: Raw
- Midjourney: Quality
- Midjourney: Stylize
- Midjourney: Chaos / Variety
- Midjourney: Weird
- Midjourney: Aspect Ratio
- Midjourney: No
- Midjourney: Image Prompts
- Midjourney: Style Reference
- Midjourney: Upscalers
- Midjourney: Zoom Out
- Midjourney: Omni Reference
Build your own prompt
Describe your image and copy it formatted for Midjourney, Stable Diffusion, Flux, ChatGPT or Gemini.
Open the prompt generator