The Complete Guide to AI Image Prompts for Beginners

Writing AI image prompts is a skill. Beginners get muddy, generic results. People who understand prompt structure get consistent, professional images. This guide covers everything.

The Anatomy of a Great Image Prompt

A high-quality image prompt has five layers:

1. Subject — What is in the image
2. Style — What it looks like aesthetically
3. Lighting — How it is lit
4. Composition — How it is framed
5. Technical specs — Resolution, aspect ratio, model parameters

Most beginners only write the subject. Professionals write all five.

Layer 1: Subject Description

Be specific about your subject. Instead of "a woman," write: *"a 30-year-old woman with short brown hair, wearing a navy blazer, sitting at a modern desk"*

Include: - Physical description - Clothing and accessories - Action or pose - Facial expression (if relevant) - Any objects in the scene

Layer 2: Style

Choose a visual style and be explicit: - "photorealistic, DSLR photography" - "oil painting, impressionist style" - "flat vector illustration, minimal" - "3D render, Pixar style" - "cinematic, film photography" - "anime, Studio Ghibli style" - "watercolor painting, loose brushwork"

Layer 3: Lighting

Lighting transforms images more than almost any other factor: - "golden hour lighting" — warm, directional, cinematic - "studio lighting, softbox" — professional, even, commercial - "dramatic side lighting" — moody, high contrast - "overcast natural light" — soft, diffused, realistic - "neon lighting, cyberpunk" — colored, atmospheric - "backlit, silhouette" — mysterious, dramatic

Layer 4: Composition

Tell the AI how to frame the shot: - "close-up portrait, face filling frame" - "wide establishing shot, full scene" - "rule of thirds composition" - "overhead flat lay view" - "low angle shot, looking up" - "bird's eye view" - "shallow depth of field, bokeh background"

Layer 5: Technical Specs (Midjourney)

• --ar 16:9 — widescreen
• --ar 1:1 — square (Instagram)
• --ar 4:5 — portrait (Instagram ads)
• --ar 9:16 — vertical (Stories/TikTok)
• --v 6.1 — latest model
• --style raw — less artistic, more photorealistic
• --q 2 — higher quality rendering

Full Example Prompt

*"35mm portrait photograph of a confident female entrepreneur, 40s, silver hair, wearing a white linen shirt, sitting at a sunlit cafe table with a laptop, shallow depth of field, warm golden afternoon light streaming from the left, natural authentic expression, photorealistic, editorial magazine style --ar 4:5 --style raw --v 6.1"*

Compare this to just: "woman at cafe" — the difference in output quality is enormous.

Platform-Specific Tips

Midjourney: Uses "--" parameters, excels at artistic and stylized images, best for creative work.

DALL-E 3 (ChatGPT): Write prompts as natural language sentences, not keyword lists. Very good at following specific instructions and text within images.

Stable Diffusion: Supports negative prompts — add "ugly, deformed, blurry, low quality" to negative prompt field to eliminate common artifacts.

Ideogram: Exceptional at text within images (logos, posters, signs) — the best choice when your image needs readable text.

Practice is the fastest teacher. Generate 10 variations of the same prompt with single-word changes to understand exactly how each element affects the output.