The universal prompt formula
Every strong image prompt, regardless of model, follows the same skeleton. Fill each slot and you have a professional prompt:
[subject] + [action/pose] + [environment] + [composition] + [lighting] + [color/mood] + [style/medium] + [camera/lens] + [quality] + [model parameters]
Example: a red fox, curled asleep, in a snowy pine forest, centered composition, soft morning backlight, cool blue palette, photorealistic, 85mm f/1.8, ultra-detailed --ar 3:2 --v 6
Midjourney parameters
| Parameter | What it does |
|---|---|
| --ar W:H | Aspect ratio, e.g. --ar 16:9, --ar 3:2, --ar 1:1 |
| --v 6 | Model version (latest: 6). --niji 6 for anime. |
| --stylize 0–1000 | How strongly Midjourney applies its aesthetic. Low = literal, high = artistic. |
| --chaos 0–100 | Variety across the 4 results. High = more unexpected. |
| --weird 0–3000 | Pushes toward unusual, offbeat aesthetics. |
| --quality .25–1 | Render time/detail. Lower = faster, cheaper. |
| --seed 0–4294967295 | Reproducibility. Same seed + prompt = similar image. |
| --no X | Negative prompt — exclude X (e.g. --no text, watermark). |
| --tile | Seamless repeating pattern. |
| --cref URL | Character reference for consistent characters. |
Stable Diffusion syntax
| Syntax | Effect |
|---|---|
| (word:1.3) | Increase a term’s weight (emphasis). 1.0 = neutral. |
| (word:0.6) | Decrease a term’s weight. |
| ((word)) | Shorthand emphasis (~1.21× per pair). |
| Negative prompt | Separate field — list what to avoid: blurry, lowres, extra fingers, watermark, text. |
| CFG scale 5–12 | How strictly it follows the prompt. ~7 is balanced. |
| Steps 20–40 | Sampling steps. More = cleaner, slower. |
| Sampler | DPM++ 2M Karras / Euler a are common defaults. |
DALL-E 3 & Flux tips
- DALL-E 3: prefers full natural-language sentences over keyword lists. Describe the scene like you would to a person. No parameter syntax.
- DALL-E 3: excellent at rendering text in images — quote the exact words: a sign that reads "OPEN".
- Flux: handles long, detailed natural-language prompts well and renders text reliably. Be specific about composition and lighting.
- Flux: no negative prompt — phrase positively (say what you want, not what you don’t).
Lighting keywords
| Keyword | Look |
|---|---|
| golden hour | Warm, soft, low-angle sunlight. |
| rim / backlight | Glowing outline separating subject from background. |
| Rembrandt lighting | Dramatic portrait light, triangle on the cheek. |
| softbox / studio | Even, flattering, controlled. |
| volumetric light | Visible light rays / god rays. |
| neon / cyberpunk | Saturated colored glow. |
Camera & lens keywords
| Keyword | Effect |
|---|---|
| 85mm f/1.4 | Portrait compression + creamy bokeh. |
| 35mm | Natural, environmental framing. |
| macro | Extreme close-up detail. |
| wide-angle 24mm | Expansive scenes, slight distortion. |
| shallow depth of field | Blurred background, sharp subject. |
| aerial / drone shot | Top-down or high vantage. |
Don’t write prompts from scratch
Upload any image and PromptShotAI reverse-engineers the full prompt for you — free.
Try the free Image-to-Prompt tool →Free to reference and share. Found it useful? A link back to PromptShotAI is appreciated.