Free Guide
How to Write AI Image Prompts That Actually Work
A practical guide for Midjourney, FLUX, Stable Diffusion, and ChatGPT Image
Why most AI image prompts fail
Vague subjects
Prompts like "a person in a forest" give the model almost no signal. You need to specify who the person is, what they are doing, their posture, expression, and relationship to the environment. The more specific the subject, the more consistent the output.
Missing style direction
Without a clear style reference — cinematic, painterly, photorealistic, illustration — the model defaults to a generic blend. Every strong prompt anchors itself in a visual tradition: "oil painting", "shot on 35mm film", "Ghibli-inspired", "clinical product photography".
Ignoring platform-specific parameters
Midjourney uses --ar, --style, --v flags. FLUX responds to natural prose. Stable Diffusion weights keywords with colons. Using the wrong syntax for your tool means losing control over the output entirely. Each platform has its own language — and the best results come from speaking it fluently.
Midjourney prompt structure
Midjourney responds best to a structured description followed by parameter flags. The formula is:
[subject]
+
[environment]
+
[lighting]
+
[mood]
+
[style]
+
[parameters]
Portrait
[SUBJECT] seated by a rain-streaked window, soft diffused daylight, melancholic and introspective mood, shallow depth of field, cinematic colour grade, film grain --ar 3:4 --style raw --v 6.1
Landscape
[SUBJECT] standing at the edge of a volcanic crater at dawn, ochre and magenta sky, long dramatic shadows, wide establishing shot, photorealistic atmosphere --ar 16:9 --style raw --v 6.1
Product shot
[SUBJECT] on a minimal white surface, soft studio lighting from above left, clean commercial aesthetic, sharp focus, white background, product photography --ar 1:1 --style raw --v 6.1
FLUX prompt structure
FLUX 1.1 Pro is trained on natural language — it prefers complete descriptive sentences rather than keyword lists. Write it like you are describing the scene to a photographer.
Portrait example
[SUBJECT] photographed in warm golden hour light, standing in a narrow cobblestone alley in southern Europe. The shot is taken at eye level with a 50mm lens, shallow depth of field blurring the stone walls behind. The mood is warm and contemplative, colours leaning amber and terracotta.
Landscape example
A sweeping aerial view of [SUBJECT] moving through a dense pine forest in winter. Light snow dusts the canopy. The image is captured from directly above, giving the forest a graphic, abstract quality. Muted greens and whites, overcast diffused light, no harsh shadows.
The [SUBJECT] variable technique
Instead of writing a one-off prompt for a specific image, you can write a reusable template where the primary subject is a swappable variable. This lets you reuse the same lighting, mood, camera angle, and style across dozens of different subjects without rewriting the whole prompt.
Replace [SUBJECT] with any noun and the prompt still reads naturally and produces a coherent result.
Example 1 — same template, different subjects
Template: [SUBJECT] in ancient ruins at golden hour, long amber shadows, cinematic wide shot, 85mm lens --ar 16:9 --style raw --v 6.1
With "A lone warrior": A lone warrior in ancient ruins at golden hour...
With "A wandering monk": A wandering monk in ancient ruins at golden hour...
With "An abandoned cart": An abandoned cart in ancient ruins at golden hour...
Example 2 — product template
Template: [SUBJECT] on brushed concrete, dramatic side lighting from the left, deep shadows, commercial photography, 4:5 ratio --ar 4:5 --style raw
With "A glass perfume bottle": A glass perfume bottle on brushed concrete...
With "A leather wallet": A leather wallet on brushed concrete...
Prompt templates you can use right now
Copy any of these and replace [SUBJECT] with your own subject. Each template works across multiple platforms.
Portrait — Midjourney
[SUBJECT] in soft natural window light, slight overcast diffusion, intimate close-up, shallow depth of field, film grain, muted colour palette, introspective mood --ar 3:4 --style raw --v 6.1
Landscape — FLUX 1.1 Pro
[SUBJECT] at the edge of a misty mountain valley just after sunrise. The scene is photographed from a low angle looking upward, with golden light breaking through low cloud. Colours are warm amber and cool blue-grey. Photorealistic, highly detailed atmospheric haze.
Product — Stable Diffusion
masterpiece, best quality, product photography of [SUBJECT], clean white studio background, (soft box lighting:1.3), sharp focus, commercial aesthetic, 8k, no watermark
Architecture — ChatGPT Image
A detailed architectural photograph of [SUBJECT] at blue hour, warm interior lights glowing from within, exterior lit by cool ambient dusk sky. Shot from street level with a wide-angle lens, symmetrical composition, no people.
Abstract — Midjourney
Abstract representation of [SUBJECT], fluid organic forms, deep indigo and gold colour palette, macro photography aesthetic, high contrast, dreamlike textures, no text --ar 1:1 --style raw --v 6.1
Generate your own prompt template
IdeaToImagePrompt chats with you to understand your image idea, then builds a custom reusable template — using live documentation for your specific tool.
Try it free →