Midjourney 8.1: how to write prompts the model actually understands

What changed in V8.1

V8.1 is not a separate product — it's an upgrade to V8. But Midjourney removed several parameters that used to work, which is critical for prompt scoring.

Removed: `--no` (negatives — describe positively now), `--ow` (omni-weight — no replacement), `--p 0` (disabling personalization — UI only now). Midjourney's error message: `--no is not compatible with --version 8.1`. If a V8.1 prompt contains `--no`, `--ow`, or `--p 0`, that's mistake #1 for immediate flagging.

HD is default — the `--hd` flag is no longer needed; native 2K is included by default. Image references load only via 3 UI slots: Start Frame, Image Prompts, Style References. The safe set of working parameters is `--raw`, `--s`, `--ar`, `--c`, `--q`.

• Removed: --no, --ow, --p 0 (do not work in V8.1)
• HD by default — the --hd flag is gone
• Image references only via UI slots, not parameters
• Web only (alpha.midjourney.com) — Discord not supported
• Anti-bokeh stack required for documentary realism

Prompt structure

V8.1 rewards coherent cinematic descriptions, not tag lists. A 60–140-word paragraph that reads like a director briefing a cinematographer.

Element order: 1) Shot type + subject («Candid iPhone close-up of a woman in her 20s…»), 2) Specific physical details — what makes her her, not a generic «woman», 3) Wardrobe + props — materials, condition, wear, 4) Action / expression — a concrete moment, 5) Camera angle and lens feel — one short technical anchor, 6) Light + setting — source, direction, atmosphere, 7) Finishing texture — skin / grain / realism cue at the end.

After the prompt — a separate parameter block for V8.1. Core rules: state lighting explicitly, write coherent descriptions, add camera and lens, lock `--ar` from the start.

Parameter defaults by category

Ready-made baselines (Dan Kieft v3): Photorealism / documentary / cinematic action — `--raw --s 100 --ar 16:9`. Editorial portraits / fashion — `--s 100 --ar 16:9` without `--raw`. Stylized illustration / anime / cyberpunk — `--s 250 --ar 16:9` without `--raw`. Movie posters — `--s 250 --ar 2:3` without `--raw`. Fantasy character close-ups (grounded) — `--raw --s 100`; (stylized) — `--s 250` without `--raw`.

`--raw` disables Midjourney's house-style treatment (boosted contrast, warm color grade, polished glow). It makes the result more prompt-faithful. Use it for photorealism, documentary, product work. Do NOT use it for stylized illustration, fantasy art, editorial fashion, movie posters — categories that benefit from house-style.

`--s` (stylize): 100 (default) for balance, 250 for noticeably more stylized illustration, 500–750 for maximum artistic interpretation and prompt drift.

Anti-bokeh stack

V8.1 has strong baked-in bokeh — the model almost always adds shallow depth of field on close-ups. For iPhone realism or documentary photography this kills authenticity: iPhone sensors are small and naturally deep-focus; there's no real bokeh (Portrait Mode is a software fake).

Anti-bokeh stack — 3–4 phrases at once: «Deep focus, sharp from foreground to background. No blur, no bokeh, no depth-of-field separation. Shot on 24mm lens at f/11. Natural deep depth of field of an iPhone. Hyperfocal deep focus, sharp throughout entire frame.» One phrase is not enough — V8.1 will ignore it.

Critical exception: in action cinema (battle scenes, charging POV, mid-combat) shallow depth of field on the hero with soft chaos behind is correct — that's how prestige cinema actually shoots these scenes (Last Samurai, Gladiator, Pelennor Fields). Don't fight bokeh in action cinema; fight it in static realism.

Common mistakes

1. 1. Using --no in V8.1

   The `--no` parameter is removed in V8.1 — the model returns the error «--no is not compatible with --version 8.1». Replace it with positive description: instead of «--no clutter» write «clean background, minimal props». This is the most common mistake in prompts after the V8.1 update.

2. 2. Using removed --ow and --p 0 parameters

   `--ow` (omni-weight) and `--p 0` (disabling personalization) are removed in V8.1. No replacement — describe without omni-weight, manage personalization via the UI Style Creator. If copying a prompt from V8, check for these flags and strip them.

3. 3. Using --hd as a flag in V8.1

   HD is the default in V8.1; the `--hd` flag is unnecessary — native 2K is included by default. The cost has risen relative to V8 base, but that's baked into the default mode now. Specifying `--hd` won't throw an error but has no effect.

4. 4. Anti-bokeh stack in action scenes

   In action cinema (battle, mid-charge, POV) shallow depth of field on the hero is correct cinematography, not a bug — that's how prestige cinema shoots these scenes. Stacking anti-bokeh in an action scene fights the correct effect. Anti-bokeh is for iPhone realism and static documentary, not cinematic action.

5. 5. Tag-soup style (comma-separated words)

   V8.1 interprets the prompt even more literally than V8 — a list of comma-separated words produces chaos. Write coherent cinematic descriptions of 60–140 words. One paragraph, like a director briefing a cinematographer, in the order: shot type + subject → details → wardrobe → action → camera → light → texture.

Before / after examples

iPhone selfie of a woman in subway --no blur, bokeh, depth of field --v 8.1

iPhone selfie photograph of a young woman in her early 20s sitting on a bench in a busy New York City subway station, holding her iPhone at arm's length. Front camera perspective, slightly elevated angle. She wears a soft pink leotard, dark hair in a tight bun. Around her, the busy NYC subway, white tile walls, fluorescent overhead lighting. The natural deep depth of field of an iPhone front camera, everything in sharp clear focus, no blur or bokeh, no portrait mode effect. Hyper-realistic skin texture, raw unfiltered iPhone photography aesthetic. --raw --s 100 --ar 16:9

cyberpunk anime girl in neon alley

Stylized anime cyberpunk illustration in the visual style of Arcane, Cyberpunk Edgerunners, and Blade Runner Black Lotus, mid-shot of a stylish cyberpunk woman leaning against a graffiti-covered wall in a narrow rain-soaked neon alleyway, smoking a cigarette with a contemplative expression. Long straight white hair with blunt bangs, sharp angular jawline, large striking cyan eyes. Neon-lit skyscrapers, wet pavement reflecting saturated red and pink neon glow. Painterly digital illustration with visible brush strokes, anime-stylized proportions. --s 250 --ar 16:9

horror movie poster of a girl in hallway

Cinematic horror movie poster artwork, vertical composition, a small girl about seven years old walking slowly forward down the middle of an enormous long empty hallway of an old turn-of-the-century gothic mansion, viewed from far behind her at floor level. She wears a long pale Victorian-era white nightgown, dragging a worn old teddy bear. The hallway stretches into deep darkness with peeling wallpaper, dust motes in pale window light. At the very far end, a tall thin shadowy figure stands watching her. Cool desaturated palette, the look of a prestige A24 horror poster. Significant negative space at top for typography. --s 250 --ar 2:3