AI Image Prompt Guide: Write Better Prompts for Midjourney & DALL-E

The quality of an AI-generated image is almost entirely determined by the quality of the prompt. This guide teaches you the exact framework to write prompts that consistently produce stunning results in Midjourney, DALL-E, and Stable Diffusion.

The quality of an AI-generated image is almost entirely determined by the quality of the prompt. Two people with the same AI image generator can produce wildly different results based solely on how they describe what they want. This guide gives you the complete framework for writing prompts that consistently produce professional-quality results.

The Anatomy of an Effective AI Image Prompt

The best AI image prompts follow a consistent structure, regardless of which tool you're using:

[Subject] + [Style/Medium] + [Composition] + [Lighting] + [Mood/Atmosphere] + [Technical Parameters]

Each element adds a layer of control. Omit them and you're leaving quality to chance.

Element 1: The Subject

The subject is what your image is about. Be specific. The difference between "a woman" and "a 30-year-old Japanese businesswoman in a charcoal blazer, looking at her phone with a slight smile" is the difference between a generic stock photo and something usable.

Specificity tips:

Include age, appearance, ethnicity, and expression for portraits
Describe the environment for scenes (time of day, weather, indoor/outdoor)
Specify the action or pose, not just the person or object
Name specific objects and how they relate to each other

Example transformation:

Weak: "a dog on the beach"
Strong: "a golden retriever running through shallow ocean waves at golden hour, water splashing around its legs, motion blur on the water, sharp focus on the dog"

Element 2: Style and Medium

Specifying an artistic style or medium is one of the highest-leverage moves in prompt writing. It immediately constrains the aesthetic space and produces more cohesive results.

Popular style keywords:

Photography styles: "Canon EOS R5 photograph," "35mm film photography," "Polaroid photo," "macro photography"
Art styles: "oil painting," "watercolor," "digital art," "pencil sketch," "lithograph"
Design styles: "flat design," "isometric illustration," "vector art," "Art Deco"
Artist references: "in the style of [artist]" — use sparingly and check copyright implications

Medium specificity example:

Generic: "portrait of a man"
Styled: "oil painting portrait of a man, thick impasto brushstrokes, warm earth tones, Rembrandt lighting, museum quality"

Element 3: Composition

Composition tells the AI how to frame the image. Without composition guidance, AI tools default to centered, medium-shot compositions that are technically fine but rarely interesting.

Composition terms:

Distance: close-up, medium shot, wide angle, establishing shot, aerial view
Angle: eye-level, low angle, high angle, Dutch angle, bird's-eye view, worm's-eye view
Rule of thirds: "subject positioned in the left third," "negative space on the right"
Depth: "shallow depth of field," "bokeh background," "foreground element," "leading lines"

Composition example:

Generic: "coffee shop interior"
Composed: "wide-angle shot of a cozy coffee shop interior, shot from the doorway, tables receding into the background, warm morning light through large windows, slight vignetting at edges, 24mm lens"

Element 4: Lighting

Lighting transforms mood more than almost any other element. AI image generators respond extremely well to specific lighting descriptions.

Lighting vocabulary:

Natural: "golden hour," "blue hour," "overcast diffused light," "harsh midday sun," "dappled shade"
Studio: "Rembrandt lighting," "three-point lighting," "rim lighting," "butterfly lighting," "loop lighting"
Cinematic: "anamorphic lens flare," "chiaroscuro," "motivated lighting," "practical lighting"
Specific quality: "soft diffused light," "harsh directional light," "ambient occlusion," "global illumination"

Lighting transformation:

Before: "portrait of a chef"
After: "portrait of a chef in a professional kitchen, dramatic Rembrandt lighting from the left, warm tungsten light from the stove, rim light separating subject from background, moody atmospheric"

Element 5: Mood and Atmosphere

Mood words are surprisingly powerful in AI prompts. They activate associations that influence color palette, contrast, and overall feel without explicitly specifying each parameter.

Mood vocabulary:

Emotional: "melancholic," "joyful," "tense," "peaceful," "nostalgic," "ethereal"
Aesthetic: "cinematic," "editorial," "commercial," "fine art," "documentary"
Atmosphere: "hazy," "crisp," "saturated," "desaturated," "high contrast," "flat"

Midjourney-Specific Parameters

Midjourney uses parameter flags appended to your prompt:

--ar [ratio] — Aspect ratio (e.g., --ar 16:9 for widescreen, --ar 2:3 for portrait)
--v [version] — Model version (--v 6 is current best quality)
--q [1-2] — Quality level (--q 2 for maximum detail)
--s [0-1000] — Stylize amount (higher = more artistic interpretation)
--chaos [0-100] — Variation between results
--no [element] — Negative prompt to exclude elements (e.g., --no text, watermark)

Full Midjourney prompt example: "Editorial fashion photography, female model in flowing silk dress, rooftop at sunset, Tokyo skyline in background, warm golden light, shallow depth of field, Vogue Magazine aesthetic, high fashion, cinematic --ar 2:3 --v 6 --q 2 --s 250"

DALL-E 3 Specific Tips

DALL-E 3 (used in ChatGPT) processes natural language more intuitively than Midjourney. Key differences:

More conversational prompts work better — full sentences with context, not comma-separated keywords
It follows instructions literally — if you say "a red ball on the left," it will precisely place a red ball on the left
Text generation is much better — DALL-E 3 can render readable text in images, which Midjourney struggles with
Style references work differently — describe the style outcome you want rather than naming artists

DALL-E 3 prompt example: "Create a warm, inviting photograph of a modern home office at dusk. The desk faces a large window overlooking a garden. Soft lamp light illuminates the desk surface where an open laptop sits next to a coffee mug and a small potted succulent. The atmosphere is cozy and productive, with a slightly warm color temperature and gentle shadows."

Common Prompting Mistakes

Mistake 1: Being too short. "A sunset" gives the AI too much creative latitude. You'll get a generic sunset. Add details about what makes this sunset specific.

Mistake 2: Conflicting instructions. "Hyper-realistic AND cartoon-style" creates visual confusion. Pick one aesthetic direction.

Mistake 3: Neglecting negative prompts. Most tools support negative prompts. Use them to exclude: "blurry, low quality, watermark, text, extra fingers, deformed" prevents common AI artifacts.

Mistake 4: One-shot giving up. The best results come from iteration. Generate 4 results, identify the best one, then use Midjourney's "vary subtle" or DALL-E's "edit this image" to refine rather than starting over.

Building a Prompt Library

As you find prompts that work, save them. A good prompt for "professional headshot" or "product photography" can be reused with minor modifications across dozens of projects. Tools like Notion, Obsidian, or even a simple text file organized by category become invaluable as your prompt vocabulary grows.

The difference between professionals and beginners in AI image generation isn't access to better tools — it's having a refined library of prompts built through iteration and experimentation.

For related discovery, browse the AI Tools Directory and verify each provider's current terms before choosing a tool.