How to Write a Midjourney Prompt: From Idea to Image

What You Will Deliver

This workflow builds a complete Midjourney prompt layer by layer: subject, style, environment, lighting, camera, and parameters. By the end you will have one polished prompt you can paste into Midjourney and a reusable formula for every image after it.

Midjourney reads prompts left to right, weighing the first words most. So we build the prompt from the most important element (the subject) outward to the least important (parameters). For version-specific behavior, see Midjourney v7 prompt changes — the formula below works on v6 and v7.

Step 1: Define the Subject

Start with a concrete noun phrase. “A cat” is too weak; “a ginger tabby cat wearing a knitted scarf” gives the model something to anchor on.

Subject = a ginger tabby cat wearing a knitted blue scarf
Write 5 alternative subject lines for this idea, each adding a specific detail (breed, clothing, pose, or expression). Return one per line, no explanation.

Pick the subject you like. This is the heaviest-weighted part of the final prompt, so spend the most effort here.

Step 2: Choose the Style

Style tells Midjourney the visual language: photorealistic, oil painting, pixel art, cyberpunk render, etc. Combine an art medium with a reference artist or movement for stronger results.

Style = cinematic photorealistic, 35mm film look, inspired by Wes Anderson color palette
Give me 4 style descriptors I can append to a prompt for "[SUBJECT]". Each should name a medium (photo / illustration / 3D render) and a mood. One per line.

Append your chosen style after the subject with a comma. Keep it to two or three descriptors — too many styles fight each other.

Step 3: Set the Environment

The setting grounds the image. Indoor, outdoor, weather, time of day, and background detail all matter.

Environment = on a snowy Paris balcony at dawn, blurred city behind
Suggest 3 environment phrases for "[SUBJECT]" that change the mood (cozy, lonely, adventurous). One phrase per line.

Place the environment right after the style. Midjourney uses it to fill the frame around your subject. A useful trick is to pair the environment with a weather or time cue, because those change the whole mood cheaply: “at dawn” reads calm and fresh, “midnight rain” reads moody and noir, “noon market square” reads busy and energetic. Pick the cue that matches the feeling you want before moving on.

Step 4: Control Lighting

Lighting is the cheapest way to make an image look professional. Name the source, direction, and quality.

Lighting = soft golden-hour side light, gentle rim light, shallow depth of field
List 4 lighting setups (e.g. hard noon sun, neon backlight, studio softbox, candlelit) for "[SUBJECT]". One per line, each under 8 words.

Add lighting after the environment. Pair it with the camera setting in the next step for cohesive results.

Step 5: Pick the Camera and Parameters

Camera words (lens, angle, film stock) plus Midjourney parameters (–ar, –v, –style) finalize the look. Assemble everything into one prompt.

Combine these into one Midjourney prompt, comma-separated, then append parameters:
Subject: [YOUR SUBJECT]
Style: [YOUR STYLE]
Environment: [YOUR ENVIRONMENT]
Lighting: [YOUR LIGHTING]
Camera: 50mm lens, eye-level angle, shot on Kodak Portra 400
Parameters: --ar 3:2 --v 7 --style raw
Return only the final prompt on one line, ready to paste.

Your final prompt looks like: a ginger tabby cat wearing a knitted blue scarf, cinematic photorealistic, on a snowy Paris balcony at dawn, soft golden-hour side light, 50mm lens, eye-level –ar 3:2 –v 7 –style raw. For product-style shots, the approach in Midjourney product mockup prompts reuses this same assembly with a clean-background variant. If you need the same character across many images, the character consistency prompts extend this formula with a reusable character sheet.

A Full Worked Example

Let’s assemble a second image so the pattern is concrete. Subject: “a leather messenger bag on a wooden table.” Style: “studio product photography, soft editorial look.” Environment: “near a sunlit window, plant in background, shallow depth of field.” Lighting: “diffused window light, gentle highlights on the leather.” Camera: “85mm macro, top-down 30-degree angle.” Parameters: “–ar 4:5 –v 7 –style raw.” The pasted prompt becomes: a leather messenger bag on a wooden table, studio product photography, near a sunlit window with a plant behind, diffused window light, 85mm macro top-down –ar 4:5 –v 7 –style raw. Notice the subject leads, the descriptive layers follow in weight order, and parameters sit last where Midjourney treats them as settings rather than scene content.

Final Deliverable Checklist

Tips for Consistent Results

Once you like a result, save the prompt as a reusable template and only swap the subject. This keeps your style stable across a series of images, which matters if you are building a brand or a story. Another habit: generate a small batch (the –repeat or variation feature) from one prompt, then pick the best rather than tweaking endlessly — Midjourney surprises you more than micro-editing does. If a concept will not render, describe it with concrete visual words instead of an abstract name; “a cozy room” beats “hygge,” which the model may not parse.

Common Pitfalls

1. Writing the prompt backwards (parameters first). Weight flows left to right; put the subject first.

2. Stuffing 10 styles. They cancel out. Two or three is plenty.

3. Forgetting aspect ratio. Default is square; set –ar for the platform you need.

4. Using vague words like “beautiful” or “nice.” Midjourney needs concrete visual terms.

Sources & Notes

Prompt formula validated with Midjourney v6.1 and v7 in October 2026. The left-to-right weighting and parameter syntax (–ar, –v, –style) reflect current Midjourney documentation; parameter names may shift between major versions, so always confirm against the in-app /help for your version. Category mapping follows the guunn.com AI Prompts plan (Midjourney = 958). No images were generated and no APIs were called during article creation; prompts are copy-paste text for the reader.