Midjourney No Parameter (--no): The Exclusion Reference (2026)

TL;DR: Midjourney's --no parameter is a bias, not a filter. Its own docs state that --no is identical to applying a -0.5 weight to that term in a multi-prompt, so it discourages an element rather than removing it. It goes at the end of the prompt and accepts comma-separated terms. Checked August 27, 2026.

What does Midjourney's --no parameter actually do?

It lowers the weight of a term. It does not delete it.

The --no article headlines itself "Tell Midjourney what you don't want in your image using the no parameter". That framing is where most guides stop, and the mechanism is published on a different page.

Midjourney's Multi-Prompts and Weights article, which defines the :: weighting syntax, states that using the no parameter "is the same as using a -0.5 weight." It prints the equivalence itself, giving vibrant tulip fields --no red and vibrant tulip fields:: red::-0.5 as the same prompt. The --no article agrees, describing the parameter as "the same as weighing part of a multi-prompt to" a value of -0.5.

So this:

sunlit scandinavian kitchen, morning light --no plants

is documented as equivalent to this:

sunlit scandinavian kitchen, morning light:: plants::-0.5

Three consequences follow, and every one of them explains a common complaint.

A negative weight is finite. -0.5 is a specific number, and the rest of your prompt carries a default weight of 1. A prompt heavy with greenhouse language will still produce plants, because you asked for them harder than you unasked for them.

It is relative, not absolute. The weight competes against everything else in the prompt, so adding more scene description changes what -0.5 is measured against.

Nothing promises removal. Midjourney publishes no floor and no statement about what happens when the model cannot honour the bias. If an element absolutely must not appear, --no is the wrong tool.

Why doesn't "without any fruit" work in a Midjourney prompt?

Because the model reads it as a description containing fruit.

Midjourney publishes its own worked example of this failure, which is unusual and worth using. Its --no article prints two prompts under a thumbs-down icon: "still life gouache painting without any fruit" and "still life gouache painting, please don't add fruit!" Under a thumbs-up icon it prints one: "still life gouache painting --no fruit, apple, pear".

The Prompt Basics article gives the general rule with a different example, dated March 24, 2026, telling you to describe what you do want instead of what you do not. If you mention a party with no cake, it says, "a cake might still appear."

The practical rule is that negation belongs in the mechanism, never in the sentence. Prose describes the scene. The parameter adjusts weights.

Does --no work in Midjourney V8.2?

Midjourney publishes two answers, and they are not compatible with each other.

The --no article carries a compatibility banner reading "This feature is supported in V8.2". The Feature Compatibility chart inside the Version article gives the No Parameter its own row, with a check icon in every column including V8.1 and V8.2.

The Multi-Prompts and Weights article, which is the page that defines what --no does, carries a banner reading "This feature is not supported in V7 or V8.2". Its body narrows further, saying multi-prompts work with Midjourney versions "1, 2, 3, 4, Niji 4, 5, Niji 5, 6, Niji 6, and 6.1." That list omits V7 and the entire V8 series. The same compatibility chart marks Multi-Prompting with the no-symbol icon in both the V7 and the V8.1 and V8.2 columns.

Two rows of one table, four rows apart, saying that a flag is supported while the mechanism it is defined as being identical to is not.

This is not the only place the chart and the prose diverge. The Quality article discusses only V7 while the chart marks --quality unsupported on V8.1 and V8.2, which is covered in the quality and speed reference. The Legacy Features article, updated July 9, 2026, still tells readers to see the Version article "For information on the current default version (V7)" more than three weeks after V8.1 became default. Where a Midjourney parameter article and the compatibility chart disagree, check the update date on both, then check your own settings panel.

V8.2 has been the default since July 24, 2026, per the Version article: "V8.2 released as the default version on July 24, 2026." If you are unsure which model you are actually on, the version flag reference covers --v and --niji.

Can you put multiple terms in one --no?

Yes, per the --no article, and no, per the Parameter List. Both are live.

The --no article is explicit and gives a worked example: "You can even list multiple elements by separating them with commas." Its good prompt is still life gouache painting --no fruit, apple, pear.

The Parameter List article, last updated June 11, 2026, lists three formatting rules for every parameter. The third reads: "Don't use commas, periods, or other punctuation marks in your parameters." Its illustrated bad example is a trailing comma after --ar 2:3, labelled "Using punctuation in parameters", not a --no list. But the rule as written is general.

The rest of the Parameter List's formatting rules do apply, and they are the first thing to check when --no appears to do nothing at all. "Always put parameters after your prompt text." Put a space before the dashes. Do not put descriptive text after the parameter block.

# Wrong: prompt text after the parameter block
a wide desert highway --no cars, powerlines at golden hour

# Right: all descriptive text first, parameters last
a wide desert highway at golden hour --no cars, powerlines

How negative can a Midjourney weight go?

Further than -0.5, but only through :: syntax, and only until the sum goes negative.

Midjourney's Multi-Prompts article publishes the constraint directly: "the total of all weights in your prompt needs to be a positive number." It then works the arithmetic both ways. still life painting:: fruit::-0.5 is fine because the first phrase carries a default weight of 1 and 1 plus -0.5 is 0.5. But still life painting:: fruit::-2 "won't work because 1 plus -2 equals -1, and that's not positive. If it's negative, Midjourney will show you an error."

So the ceiling on how hard you can push against a term is set by how much positive weight the rest of your prompt carries. That is the entire published theory, and there is nothing in it about --no specifically, because --no is locked to one value.

The catch is the version story above. On V6 and earlier you can express a stronger negative with :: weights, because Multi-Prompting is marked supported there. On V7 and the V8 series it is marked unsupported, which leaves --no as the only exclusion surface, fixed at one value. Midjourney does not publish a workaround, and neither should anyone else.

Does --no interact with --sref, --stylize or image prompts?

Not in any documented way. That is a real answer, not a dodge.

I searched all 105 articles in Midjourney's help centre on August 27, 2026 for the token --no. It appears in exactly three: the No article itself, one sentence of Multi-Prompts and Weights, and one table cell of the Parameter List. It does not appear in Style Reference, Stylize, Image Prompts, Chaos, Raw, Omni Reference, Moodboards or Personalization.

So there is no published statement that --no competes with a style reference, that a high --stylize value overrides it, or that an image prompt outweighs it. The Stylize article documents only its own range: the default is 100, adjustable between 0 and 1000 on current versions. Anyone telling you that --stylize 750 cancels --no is reporting a personal impression. It may well be right, but it is not documentation.

What Midjourney does publish is where --no is compatible with its editing tools, in two lists that are easy to miss.

The video exclusion is the firmest of these. The Video article states that "Video generations are only compatible with video-specific parameters:" and names --motion low, --motion high, --raw, --loop, --end and --bs. It adds that any image parameters "used to generate your original image will be automatically removed when generating videos." So --no shapes the still frame you animate, and stops there.

When should you re-describe the scene instead?

When absence matters more than emphasis, and when the term is more than one word.

The second case is the sharper one, because Midjourney publishes the failure mode. Its moderation system, the --no article says, "reads every word you add to the" parameter independently. The example it gives is that --no modern clothing is read as "no modern" and "no clothing", which "can accidentally trigger a warning". Its own recommended fix is not a better negative: "include the type of clothing you do want in your prompt, rather than using" --no.

That advice generalises. Three situations where re-describing beats excluding:

The element is structural. "No people" fights the entire concept of a street scene. "An empty pedestrianised street at dawn, shutters down" gives the model a coherent thing to render instead of a thing to avoid.

The negative is a compound phrase. Anything of the form adjective plus noun gets split into independent tokens. --no modern furniture is two negatives, not one, and one of them is the word "modern".

You are excluding a style rather than an object. "Not cartoonish" has no object to bias against. Naming the medium works better: gouache, tintype, architectural photography. The common Midjourney prompt failures are mostly of this shape.

The reliable pattern is two-stage. Describe the scene specifically enough that the unwanted element has no reason to exist, then use --no for the one or two stragglers that keep appearing anyway. Keeping that structure as a saved prompt template, with the subject and the exclusion list as variables, outlasts any single parameter value, because this surface moved four times in 2026 alone.

18 copy-paste Midjourney --no prompts

Every prompt below puts descriptive text first and parameters last. Swap the subject freely; the structure is the point. Add --v 8.2 or your preferred version flag if your default is set elsewhere, and see the aspect ratio reference for --ar values.

Single-term exclusions, the safest form.

# 1. One word, one concept, no compound phrase to split.
sunlit scandinavian kitchen, oak counters, morning light --ar 3:2 --no plants

# 2. Excluding a colour, the example Midjourney itself uses.
vibrant tulip fields under a wide overcast sky --ar 16:9 --no red

# 3. Excluding an object class from an interior.
a reading nook with a linen armchair and a brass floor lamp --ar 4:5 --no books

# 4. Excluding a texture rather than an object.
a matte ceramic vase on a pale plaster shelf --ar 1:1 --no gloss

Comma-separated lists, per the No article's own example.

# 5. Midjourney's documented good prompt, verbatim.
still life gouache painting --no fruit, apple, pear

# 6. Product photography without the usual props.
a walnut pour-over stand on a concrete counter, soft window light --ar 4:5 --no hands, steam, plants

# 7. Architectural exterior with the clutter stripped out.
a brutalist library facade at dusk, long exposure --ar 21:9 --no cars, people, signage

# 8. Landscape with the modern world removed.
a chalk stream winding through downland, early autumn --ar 3:2 --no fences, powerlines, roads

Editorial and portrait work.

# 9. Studio portrait without the fashion-shoot defaults.
studio portrait of a ceramicist, 85mm, single softbox, grey seamless --ar 4:5 --no jewelry, glasses

# 10. Documentary feel rather than commercial.
a fishmonger at a covered market, available light, film grain --ar 3:2 --no smile, logos

# 11. Excluding the AI-portrait house style.
a weathered lighthouse keeper, overcast daylight, tintype --ar 4:5 --no bokeh, lens flare

Illustration and graphic work.

# 12. Flat vector without the depth cues.
flat vector illustration of a mountain rail line, four-colour palette --ar 16:9 --no gradient, shadow

# 13. Line art with nothing filled in.
technical line drawing of a bicycle drivetrain, ink on cream paper --ar 1:1 --no color, shading

# 14. Pattern work that must stay clean.
seamless botanical pattern, two-tone sage and bone --ar 1:1 --tile --no text, border

The traps, written out so you can see them fail.

# 15. WRONG. Compound negative: parsed as "no modern" plus "no clothing".
a portrait of a folk musician on a village green --no modern clothing

# 16. RIGHT. Same intent, expressed positively, per Midjourney's own advice.
a portrait of a folk musician on a village green, wool waistcoat, linen shirt, 1890s dress

# 17. WRONG. Negation in prose. The model reads "cake" as part of the scene.
a birthday party in a garden with no cake --ar 3:2

# 18. RIGHT. Describe the scene you want, then exclude the straggler.
a garden gathering with bunting, folding chairs and a long trestle table --ar 3:2 --no cake

What Midjourney does not publish about --no

Writing the gaps down is more useful than filling them with plausible-sounding numbers, because every item here is currently answered with confidence somewhere on the open web.

• What --no does on V8.1 and V8.2. The flag is marked supported; the -0.5 multi-prompt weight it is defined as is marked unsupported on the same chart. No page reconciles them.
• Whether repeated --no flags stack. Two --no blocks in one prompt is not documented in any direction.
• Any interaction with --sref, --stylize, --iw or --oref. Zero mentions across all 105 help articles.
• Whether --no affects upscales, Pan or Zoom Out. Documented for Vary Region and Remix only.
• A maximum number of comma-separated terms. No cap is published.
• What happens when the bias fails. No error, no warning, no fallback behaviour is described.
• Whether the moderation word-splitting applies to every term in a list or only to multi-word terms.

Everything above is a genuine gap in the vendor documentation as of August 27, 2026, not a gap in this article. Community consensus exists for several of them and is worth reading, but it is not the same kind of claim and should not be quoted as though a vendor said it.

How to verify all of this yourself

The docs move, so the durable skill is checking rather than remembering.

Read the compatibility chart, not the parameter article. The Feature Compatibility chart in the Version article is the only place per-version availability is stated for the No Parameter, Multi-Prompting, Quality and Turbo side by side. Where it and a parameter article disagree, compare their update dates.

Fetch the docs the way that works. docs.midjourney.com returns 403 to most automated requests, which is not a dead link and does not need fixing. The Zendesk Help Center JSON API behind it answers plainly at /api/v2/help_center/en-us/articles/<id>.json, and updates.midjourney.com serves the dated changelog posts that the help centre summarises later and occasionally contradicts. The No article is 32173351982093, Multi-Prompts and Weights is 32658968492557, and Version is 32199405667853.

Test the equivalence yourself. Run the same prompt three ways on a fixed seed: once plain, once with --no term, and once with :: term::-0.5. On a version where multi-prompting is marked supported, those last two are documented as the same prompt. On a version where it is marked unsupported, whatever you observe is the only evidence there is, because Midjourney has not written it down.

Every figure, quote and date on this page was read from the Midjourney help centre or its changelog on August 27, 2026. Nothing here came from a search summary or a secondary guide.