home/image-gen/prompt-structure

Prompt Structure

GPTClaudeGemini··1,170 copies·updated 2026-07-14
prompt-structure.prompt
# Prompt Structure Reference

The full craft behind the 10-slot spine. Read this for any non-trivial prompt.

## Contents
- [The 10 slots, with good vs weak phrasing](#the-10-slots-good-vs-weak)
- [Signature techniques (the moves that separate pro prompts)](#signature-techniques)
- [Template A — Photographic editorial](#template-a--photographic-editorial)
- [Template B — Layout / landing-page / poster](#template-b--layout--landing-page--poster)
- [Annotated walk-throughs](#annotated-walk-throughs)
- [Fill-in-the-blank skeletons](#fill-in-the-blank-skeletons)
- [Refine mode — upgrading a prompt the user brings](#refine-mode)

---

## The 10 slots, good vs weak

The spine is: **Concept → Canvas → Spatial structure → Subject → Lighting → Composition → Palette → Type/Camera → Finish → Reference stack.** Slots 3 and 8 fork by sub-type (photographic vs layout).

| Slot | Weak | Strong |
|------|------|--------|
| **1. Concept line** | "A cool photo of a snowboarder." | "Contemporary commercial lifestyle photograph, luxury snow-sports campaign aesthetic, authentic iPhone-shot energy. The creative heart of the image: a selfie reflected in a snowboarder's mirrored goggles." |
| **2. Canvas/format** | "Make it tall." | "Vertical 9:16 mobile composition. Full-bleed deep royal-blue→navy gradient (#2E5BAC top → #1A2D4F bottom), subtle film grain, slight edge vignette." |
| **3. Spatial structure** | "Person in the middle, stuff around them." | Photo: "Foreground hero …; the reflection (focal element) …; midground slope …; background alpine peaks." Layout: "Top banner …; hero photo fills ~60% …; bottom content block …; thin bottom strip." |
| **4. Subject detail** | "A stylish young guy." | "Mid-twenties man, light stubble, relaxed knowing half-smile; ski helmet + oversized chrome-mirrored goggles (silver-blue lens filling most of the frame); electric-orange technical puffer; cream gaiter under chin." |
| **5. Lighting** | "Nice lighting." | "Golden-hour alpine sun from camera-right; warm rim light wrapping the face and jacket; white snow as a giant natural bounce fill; high-key exposure; crisp high-altitude light quality." |
| **6. Composition** | "Centered." | "Portrait orientation, head-and-shoulders, goggles on the upper third occupying ~40% of the frame as the visual hero; generous breathing room; slight handheld imperfection." |
| **7. Palette** | "Colorful." | "Electric orange (#FF6A1A) jacket against deep cobalt sky (max complementary contrast); hot-pink reflection accent; chrome silver-blue lens; pure snow whites; warm skin tones." |
| **8a. Typography** (layout) | "Big bold text." | "Massive chunky slab-serif wordmark (Bungee Slab feel) for the title; bold condensed uppercase grotesk (Druk Wide) for the headline; italic serif (DM Serif Text) for body copy; small tracked caps for meta." |
| **8b. Camera** (photo) | "Shot on a good camera." | "Shot on iPhone 16 Pro main 48MP fusion at 24mm-equiv, f/1.78, ProRAW; Smart HDR 5 + Deep Fusion detail; clean digital clarity, no film grain." |
| **9. Finish** | "Make it pretty." | "Subtle film grain, slight CRT warmth, vintage screen-print feel; sticker drop-shadows on overlay elements." |
| **10. Reference stack + quality tag** | "Professional quality." | "Cass Bird × Jack Davison commercial-campaign style, Moncler × Arc'teryx ski-campaign aesthetic, Monster Children editorial quality. Awwwards/Behance-quality, 4k." |

**Length:** these prompts are long on purpose (150–400 words). For image models, dense specificity beats brevity — but every clause must change the picture. If a phrase wouldn't alter the output, cut it.

---

## Signature techniques

These are the moves that make a prompt reliably produce a polished result. Apply them by default.

1. **Inline hex codes.** Pin every load-bearing color with a hex value next to its name (`marigold-yellow (#F4C542)`). Named colors drift between regenerations; hex anchors them and keeps a series on-palette.
2. **%-of-frame callouts.** State how much space the hero occupies ("fills ~60% of vertical space", "goggles occupy ~40% of the frame"). This controls scale and focal hierarchy far better than "big" or "prominent".
3. **Name the conceptual hook.** If the image has a clever idea (a reflection, a double-portrait, a visual pun), call it out explicitly and label it ("the creative heart of the image", "key storytelling element"). The model protects what you name.
4. **IP-safety guards.** For any brand, character, mascot, or logo: mark it `fictional`/`placeholder`, then add `generic … NOT resembling [real brand] — an original design`. This keeps you clear of trademark issues *and* stops the model from lazily copying a known logo. Do this every time a brand-like element appears.
5. **Reference stack as a quality dial.** End with a short stack — `photographer × brand-campaign × magazine` + a quality tag (`Awwwards/Behance-quality`, `4k`). A few names borrow an entire aesthetic. Keep it to real, well-known references that actually match the intended look.
6. **Section labels.** For complex prompts, label blocks (`Canvas:`, `Lighting:`, `Palette:`, `Typography:`, `Finish:`). It keeps long prompts parseable and stops slots from bleeding together.
7. **Spatial layering (photo) / zone stacking (layout).** Always give the model a depth or reading order — foreground→background, or top→bottom — so elements don't pile into one plane.
8. **Intentional imperfection.** For "authentic" / phone-shot looks, ask for it explicitly ("slight handheld framing imperfection", "slight motion blur in one hand"). Models default to sterile perfection otherwise.

---

## Template A — Photographic editorial

Use for photos, portraits, lifestyle, product-in-scene, campaign imagery.

when to use it

Community prompt sourced from the open-source GitHub repo MD-Stellr/image-prompt-crafter (MIT). A "Prompt Structure" style prompt — adapt the placeholders and specifics to your task. Imported as-is and not independently retested here, so check the output before relying on it.

tags

image-gencommunitygeneral

source

MD-Stellr/image-prompt-crafter · MIT