The Midjourney Prompt Builder

I want to create an image using Midjourney, but I need help translating what I see in my head into a prompt that actually produces what I want. Help me build a precise, effective Midjourney prompt.

## What I want to create

**Describe the image in your own words:** [Describe what you see in your mind. Be as specific as you can -- what's in the scene, what the lighting looks like, what mood you're going for, any colors that matter. Don't worry about Midjourney syntax. Just describe it like you're telling a friend what you want. Example: "A woman in her 60s sitting on a porch at dusk, reading a book, warm lamplight from inside, fireflies in the yard, feels nostalgic and peaceful, maybe shot like a film still from the 1970s."]

**Intended use:** [What is this image for? Options: social media post / Instagram / website hero image / blog post / print (poster, flyer, business card) / album cover / book cover / product mockup / presentation slide / personal project / other: specify]

**Aspect ratio preference:** [Optional -- square, landscape (16:9), portrait (9:16), ultrawide (21:9), or "I'm not sure -- recommend based on the use"]

**Any visual references?** [Optional -- "I want it to look like Wes Anderson," "the color palette from Blade Runner 2049," "Edward Hopper's lighting," "like a National Geographic photo." If you have none, leave blank and I'll suggest options based on your description.]

## How I'll build your prompt

Based on what you've described, I'll construct a Midjourney prompt with these components, each chosen deliberately:

### 1. Subject and scene description
The core of the prompt. I'll take your plain-language description and restructure it for how Midjourney reads input -- front-loading the most important elements, using specific visual language that the model responds well to. Midjourney weighs the first few words most heavily, so composition matters.

**What I'll refine:**
- Vague descriptions → specific visual language ("a pretty forest" → "old-growth redwood forest, cathedral light filtering through canopy, fern understory")
- Implied mood → explicit lighting and color direction ("peaceful" → "soft golden hour light, warm amber tones, shallow depth of field")
- Missing spatial cues → camera position and framing ("a person at a desk" → "medium shot, eye level, person at a weathered oak desk, window camera-left")

### 2. Style and medium reference
What this image *looks like* as an artifact. Options include:

- **Photographic:** editorial photography, film photography (specify stock: Kodak Portra 400, Fujifilm Pro 400H, Kodachrome), product photography, street photography, documentary
- **Cinematic:** film still, anamorphic lens flare, specific director's visual style, specific cinematographer's lighting
- **Illustrative:** watercolor, gouache, ink wash, pencil sketch, digital illustration, editorial illustration
- **Fine art:** oil painting, acrylic, charcoal, lithograph, woodcut, specific artist's style
- **Digital/3D:** 3D render, isometric, low-poly, clay render, octane render
- **Graphic design:** flat design, risograph, screen print, vintage poster, Swiss design

I'll choose based on your description and intended use. If you mentioned a reference, I'll incorporate it here.

### 3. Technical camera parameters (for photographic styles)
When the style is photographic or cinematic, specific lens and camera language dramatically changes the output:

- **Focal length:** 24mm (wide, environmental), 35mm (natural, journalistic), 50mm (standard, human-eye), 85mm (portrait, slight compression), 135mm (telephoto, heavy compression, creamy bokeh), 200mm+ (extreme compression, flattened planes)
- **Aperture cues:** "shallow depth of field" (f/1.4-2.8), "moderate depth" (f/4-5.6), "deep focus" (f/11+)
- **Camera reference:** "shot on Hasselblad" (medium format feel), "shot on Leica M" (street, grain), "shot on Fujifilm X-T5" (clean, modern), "35mm film photography" (grain, halation)

### 4. Midjourney parameters
The technical flags that control generation:

- `--ar [ratio]` — Aspect ratio. 1:1 (square, Instagram), 16:9 (landscape, web), 9:16 (portrait, stories), 3:2 (classic photo), 4:5 (Instagram portrait), 21:9 (cinematic ultrawide)
- `--v 6.1` — Model version. 6.1 is current and best for photorealism and coherent composition. I'll specify this.
- `--s [0-1000]` — Stylization. Lower (0-250) = more literal to prompt. Higher (500-1000) = more artistic interpretation. Default is 100. I'll set based on how much creative latitude you want.
- `--q 2` — Quality. Higher values spend more compute for more detail. 2 is the max. I'll use it for print-quality output.
- `--c [0-100]` — Chaos. Higher = more variation between the four output images. Useful for exploration (30-50), less useful when you know exactly what you want (0-10).
- `--no [terms]` — Negative prompt. Keeps unwanted elements out. Common: `--no text, watermark, blurry, distorted, extra fingers, cartoon` (adjusted per style).
- `--sref [URL]` — Style reference. If you provide a reference image URL, this locks the visual style across generations. Powerful for consistency.
- `--sw [0-100]` — Style reference weight. Controls how strongly the style reference influences the output.

### 5. The assembled prompt

I'll give you the complete, copy-pasteable prompt in this format:

```
/imagine [detailed scene description], [style and medium], [camera/technical specs if applicable] --ar [ratio] --v 6.1 --s [value] --q 2 --no [exclusions]
```

### 6. Variations to try

I'll also provide 2-3 alternate versions:
- **Variation A:** Different style treatment (e.g., photographic → illustrated)
- **Variation B:** Different mood/lighting (e.g., golden hour → overcast/moody)
- **Variation C:** Different framing (e.g., wide establishing → tight detail)

Each variation will be a complete, copy-pasteable prompt.

### 7. Tips for iteration

After your first generation, I'll include brief guidance on:
- How to use `--seed` to lock a composition you like and vary the style
- How to upscale effectively (which upscaler for which purpose)
- When to use inpainting vs. regenerating
- How to use image prompts (`/imagine [URL] + [text]`) to blend references with description

## Important notes

- I will not include copyrighted character names or the names of living people in the prompt. If you want "a person who looks like [celebrity]," I'll translate that into physical descriptors and styling cues instead.
- Midjourney's terms of service apply. I'll note if your request might conflict with them.
- If your description is better suited to a different tool (DALL-E for text integration, Stable Diffusion for precise control, Flux for photorealism), I'll mention it. For a full comparison, try [The AI Image Tool Matcher](/agents/prompt-image-tool-matcher).

Now describe your image and I'll build the prompt.