AI-Visual-Prompt-Cookbook

Quick Links

Category	Good for	Start with
Photo + Doodle	Social snapshots, lifestyle scenes, playful overlays	Playful Mascot Doodle Snapshot, Cozy Bedroom Doodle Companion
Zine + Collage	Fashion posters, music visuals, maximalist editorial layouts	K-pop Apocalypse Ransom Zine, Y2K Grunge Hip-hop Cutout Poster
Type Posters	Big headline systems, loud campaign graphics, visual punch	Impact Burst Halftone Comic Poster, Neon Kinetic Typographic Poster
Travel + City	Destination posters, street scenes, urban diaries	Clean Triptych Travel Vlog Thumbnail, Tokyo Kawaii Travel Collage Poster
Editorial + Minimal	Cleaner compositions, structured layouts, quieter art direction	Tri Color Hardcut Portrait Poster, Soft Analog Future Editorial Poster
Product + Campaign	Launch ads, product heroes, retail graphics	Sunlit Coastal Product Blitz, Electric Blue Silhouette Product Launch

Why This Exists

Most AI image prompts are one-off text blobs: hard to reuse, hard to compare, and hard to iterate on. This repo takes a different approach: each visual style is distilled into a structured style.json you can drop into any LLM-based image generation workflow. Same JSON, more consistent style direction across generations.

Quick Start

1. Browse the Featured Styles, Quick Links, or All Styles.
2. Open a style folder and copy the style.json.
3. Paste the full JSON into ChatGPT, Claude, Nano Banana Pro, or any LLM-based image workflow.
4. Provide your own values for the variables declared in environment_variables, or edit a case in examples[*].values.
5. Generate the final image prompt and send it to your image model.

See the Complete Example below for a full input-to-output walkthrough.

Recommended Image Models

This workflow works best with end-to-end multimodal image models that can read long structured JSON prompts and generate the final image in one step.

• ChatGPT Images 2 (OpenAI, gpt-image-2) — strong text rendering, 2K/4K output, reasoning before generation
• Nano Banana Pro (Google, Gemini 3 Pro Image) — 4K output, multilingual text accuracy, strong subject consistency

Other multimodal LLMs that accept long JSON prompts may also work but are not the primary recommendation.

Complete Example

Input → Output Walkthrough

This example uses Mono Noir Type Portrait Poster Style.

Step 1 — The Style

Create a {ASPECT_RATIO} monochrome editorial poster in the Mono Noir Type Portrait Poster Style. Style fidelity lock: {STYLE_FIDELITY_ANCHORS}. Source content to avoid: {SOURCE_CONTENT_TO_AVOID}. Scene: {SUBJECT} {SUBJECT_ACTION} with {PRODUCT_OR_PROP} in {LOCATION}. Background elements: {BACKGROUND_ELEMENTS}. Wardrobe and styling: {WARDROBE_STYLE}. Composition: one large high-contrast black-and-white photographic subject, close crop, deep charcoal background, sparse negative space, shallow depth of field, serious noir editorial mood. If the aspect ratio is 16:9, make a landscape poster with the subject weighted to the right side and the headline block on the left. If the aspect ratio is 9:16, make a vertical poster with the headline stacked in the upper-left or middle-left field and the subject cropped large on the right or lower-right. Typography: render the exact readable lowercase headline text "{MAIN_TEXT}" as three short left-aligned lines...

Step 2 — Your Variables

SUBJECT = a tired architect with silver hair
SUBJECT_ACTION = studying a folded blueprint in a late-night pause
PRODUCT_OR_PROP = a rolled plan tube and a pencil held low
LOCATION = a dim concrete studio after midnight
BACKGROUND_ELEMENTS = soft charcoal wall gradient, blurred drafting table edge, deep empty space
MAIN_TEXT = focus / outlasts / noise.
SECONDARY_TEXT = studio log 02:14
ACCENT_SYMBOL = a tiny white plus
WARDROBE_STYLE = dark work jacket over a plain black shirt
ASPECT_RATIO = 16:9

Step 3 — The Final Prompt

Create a 16:9 monochrome editorial poster in the Mono Noir Type Portrait Poster Style. Style fidelity lock: black-and-white photographic portrait, deep charcoal background, giant lowercase left-aligned headline, first word in a white label, remaining words in white, high contrast, sparse negative space, close crop. Source content to avoid: no young woman with blunt bangs, no freckles close-up, no discipline beats procrastination text, no copied face or exact source crop. Scene: a tired architect with silver hair studying a folded blueprint in a late-night pause with a rolled plan tube and a pencil held low in a dim concrete studio after midnight. Background elements: soft charcoal wall gradient, blurred drafting table edge, deep empty space. Wardrobe and styling: dark work jacket over a plain black shirt. Composition: one large high-contrast black-and-white photographic subject, close crop, deep charcoal background, sparse negative space, shallow depth of field, serious noir editorial mood. If the aspect ratio is 16:9, make a landscape poster with the subject weighted to the right side and the headline block on the left. If the aspect ratio is 9:16, make a vertical poster with the headline stacked in the upper-left or middle-left field and the subject cropped large on the right or lower-right. Typography: render the exact readable lowercase headline text "focus
outlasts
noise." as three short left-aligned lines with tight leading; put the first headline word as black type inside a crisp white rectangular label, then set the remaining lines in heavy white type directly on the dark background. Add "studio log 02:14" only as tiny unobtrusive white microcopy if it fits. Use a tiny white plus only as a minimal typographic mark. Keep type sharp, flat, square-cornered, and massive. Photo treatment: realistic black-and-white studio photography, strong shadow falloff, visible facial or fabric texture, subtle grain, no color, no illustration, no collage, no extra panels, no logos, no watermark.

Step 4 — The Result

Copy Prompt Library

Prefer the short path? Browse the generated Copy Prompt Library for a copy-ready prompt per style. The full reusable style systems still live in each style.json.

Featured Styles

Six visual systems to start with. Every style ships as one JSON plus two preview images. Browse the complete set of 182 in the All Styles gallery below.

Package Shape

styles/<style-slug>/
  style.json          # Machine-readable prompt template
  preview-16x9.jpg    # Landscape preview
  preview-9x16.jpg    # Portrait preview

style.json v2.1 / v2.2

Every style.json is self-contained: copy the whole file into your LLM, then provide values for the variables declared in environment_variables or edit one of the examples[*].values cases.

• prompt_template is the reusable style prompt with {VARIABLE} placeholders.
• environment_variables declares every placeholder the template can use.
• Newer styles set "schema_version": "2.2" and declare only the variables their template uses (for example SUBJECT, MAIN_TEXT, DESIGN_DIRECTION); older v2.1 styles use the ten standard variables.
• examples contains ready-to-edit cases; each case stores only case_name and values.
• style_fidelity_anchors and source_content_to_avoid tell the model what to preserve and what not to copy.
• negative_prompt keeps watermarks, logos, direct source recreations, and off-style outputs away.

Rendered prompts such as prompt_9x16, prompt_16x9, or full_prompt are intentionally not stored. They are derived at generation time from prompt_template plus chosen values, so the JSON stays compact and does not drift.

Validate the library with:

python3 scripts/validate-style-json.py .

All Styles

Browse all 182 styles below.

The complete library, including the featured styles above. Parts of a series share one card. For full descriptions and all file links per style, see docs/CATALOG.md.