2026-07-24
Midjourney prompts: formulas, parameters, and examples
Midjourney prompts reward specificity. The model does not read your mind — it stacks visual tokens: subject, medium, lighting, lens, era, and mood. The best Midjourney prompts follow a repeatable formula, then tune with parameters like --ar, --stylize, and --chaos.
This guide is a practical prompts for Midjourney workbook: structure first, copy-paste examples second, troubleshooting last. For text LLM prompts, see best AI prompts and ChatGPT prompts. For image tools beyond Discord, see AI art prompts. Customize any idea in the AI prompt generator or browse the AI prompts hub.
The core Midjourney prompt formula
Use this stack left to right — most important first:
[subject], [action/pose], [environment], [lighting], [camera/lens], [medium/style], [mood] --ar [ratio] --stylize [n]
| Slot | What to write | Example tokens |
|---|---|---|
| Subject | Who or what | "elderly ceramicist," "electric cargo bike" |
| Action | What is happening | "pouring glaze," "parked in rain" |
| Environment | Where | "small Kyoto workshop," "Amsterdam canal at dusk" |
| Lighting | How it is lit | "soft window light," "golden hour backlight" |
| Camera | Framing feel | "85mm portrait," "wide angle low shot" |
| Medium/style | Art direction | "editorial photo," "1970s film still," "ukiyo-e" |
| Mood | Emotional color | "quiet," "tense," "playful" |
Prompt engineering for images is mostly subtraction: if two style words fight (e.g. "minimalist" + "baroque ornament"), pick one.
Essential Midjourney parameters
| Parameter | Purpose | Starter values |
|---|---|---|
--ar |
Aspect ratio | 1:1 social, 16:9 hero, 9:16 story |
--stylize |
How hard MJ pushes its aesthetic | 100 literal, 250–750 default taste, 1000 very stylized |
--chaos |
Variation between rerolls | 0–20 consistent, 30–80 wilder |
--seed |
Reproducible composition | same seed + prompt ≈ similar layout |
--no |
Negative prompt | --no text, watermark, extra fingers |
--cref |
Character reference | URL + weight for same face/outfit |
--sref |
Style reference | URL for color/grain/palette match |
Version flags change over time; check Midjourney docs for your workspace's current defaults. The habit matters more than memorizing every flag: ratio, stylize, negative tokens.
Best Midjourney prompts for portraits
Editorial headshot
portrait of a freelance architect in her 30s, confident relaxed expression, modern glass office background softly blurred, soft key light from left, 85mm f/1.8, natural skin texture, editorial photography --ar 4:5 --stylize 200 --no makeup overload, plastic skin
Character concept (illustration)
young ranger character sheet, three poses, forest cloak with brass clasps, ink and watercolor illustration, muted earth palette, storybook fantasy --ar 3:2 --stylize 400
Cinematic still
close-up of a violinist mid-performance, sweat on brow, single spotlight, dust in the air, 35mm anamorphic film still, shallow depth of field, moody amber grade --ar 21:9 --stylize 300
Product and commercial Midjourney prompts
Clean product hero
minimalist skincare bottle on honed travertine slab, soft diffused studio light, subtle shadow, macro product photography, neutral beige background --ar 1:1 --stylize 150 --no text, label, logo
Lifestyle context
ceramic coffee mug on oak kitchen table, morning sunlight through sheer curtain, steam rising, cozy scandinavian interior, lifestyle photo --ar 4:5 --stylize 250
Food photography
rustic sourdough loaf on linen cloth, flour dust, side window light, dark moody food photography, 50mm --ar 5:4 --stylize 200 --no hands, utensils
Landscape and environment prompts
Atmospheric landscape
coastal cliffs in fog, lone lighthouse beam, long exposure waves, muted blue hour palette, fine art landscape photography --ar 16:9 --stylize 350
Fantasy environment
floating islands above cloud sea, waterfalls between rocks, bioluminescent plants, matte painting, epic scale, god rays --ar 16:9 --stylize 600 --chaos 15
Architectural viz
mid-century modern house in desert, concrete and warm wood, pool reflecting sunset, architectural visualization, clean lines, palm silhouettes --ar 3:2 --stylize 250
Style recipes (mix and match)
| Style goal | Tokens to add |
|---|---|
| Film noir | "high contrast black and white, venetian blind shadows, 1940s" |
| Anime | "cel shading, vibrant linework, studio ghibli background painting" |
| Retro print | "risograph texture, limited color palette, grain" |
| Hyperreal | "8k detail, macro textures, studio strobe, f/8 sharp" |
| Painterly | "oil on canvas, visible brushstrokes, impressionist" |
Swap the subject line; keep the style block stable for a cohesive series.
Prompts for Midjourney beginners (simple starters)
red fox in snowfall, wildlife photography, soft overcast light --ar 16:9futuristic bicycle design sketch, industrial design render, white background --ar 4:3cozy reading nook, rainy window, warm lamp, interior design photo --ar 2:3abstract geometric poster, bauhaus colors, clean vector shapes --ar 1:1astronaut planting flag on mossy planet, cinematic sci-fi, wide shot --ar 21:9
Reroll, upscale favorites, vary (subtle) before chasing --chaos.
Troubleshooting common Midjourney issues
| Problem | Fix |
|---|---|
| Wrong aspect / cropping | Set --ar explicitly |
| Too "AI glossy" | Lower --stylize, add "natural skin texture," "film grain" |
| Extra fingers / weird hands | Crop tighter, --no extra fingers, or regenerate hands in edit |
| Unwanted text | --no text, letters, watermark, signature |
| Style drift across set | Lock --seed or use --sref / --cref |
| Muddy composition | Shorten prompt; one clear subject |
How Midjourney fits the wider prompt cluster
Midjourney is one engine in a bigger prompt for AI toolkit. Text workflows — briefs, captions, art direction notes — often start in ChatGPT; see prompt engineering. Finished images may need different tools for logos or precise text-in-image; compare generators in AI art prompts.
Related reading
- AI art prompts — DALL·E, Ideogram, Flux, and more
- Best AI prompts — text templates
- ChatGPT prompts — copy-paste for OpenAI
- Prompt engineering — reliable LLM patterns