Every product parses prompts differently, but structure travels well. Fill in this order instead of stacking adjectives blindly.
Suggested order
- Subject: who/what, count, pose or action.
- Environment: indoor/outdoor, era, tidy vs messy.
- Style: photo vs illustration vs 3D; reference artists or film stock if useful.
- Light & time: soft/hard light, golden hour, neon night.
- Lens: wide vs tele, depth of field, high/low angle.
- Negative prompts (if supported): no watermark text, no extra fingers, no heavy distortion.
Tips
- Broad first, tight second—explore composition, then iterate details.
- Run the same scaffold across models to see which fits your taste.
- Copyright & likeness: check training policies and portrait rights before commercial use.
Image gen is trial and error—structured prompts cut wasted rolls.