PromptEnhancer

Generate cinematic Flux 2 [pro] image prompts from reference images and concept descriptions. Upload a moodboard, describe a vibe, and get production-ready prompts written like a senior Director of Photography.

---

What it does

PromptEnhancer runs a two-step AI pipeline via OpenRouter:

1. Visual analysis — Upload reference images and Gemini 2.5 Flash extracts abstract aesthetic properties: colour palette, light quality, compositional energy, atmosphere, cinematic style. Subjects, faces, and locations are deliberately ignored — only the feel is captured.
2. Prompt generation — MiniMax M2.5 writes 3–6 Flux 2 [pro] prompts in the voice of a senior Director of Photography. Think Denis Villeneuve's restraint, Roger Deakins' mastery of practical light, David Lynch's quiet wrongness. Each prompt is 70–120 words, specifies a lens, depth of field, lighting quality, and a distinct camera angle.

Results appear as cards. Hover any card to Copy or Revise it — the revision panel lets you describe what should change ("make it warmer", "shift to dusk", "add rain") and rewrites only that prompt in-place.

---

Quick start

Prerequisites

• Node.js (v18 or later)
• An OpenRouter API key (both models used are available on the free tier)

1. Clone and install

git clone https://github.com/joosthel/PromptEnhancer.git
cd PromptEnhancer
npm install

2. Add your API key

Create a .env.local file in the project root:

echo 'OPENROUTER_API_KEY=sk-or-v1-...' > .env.local

Replace sk-or-v1-... with your actual key from openrouter.ai/keys.

3. Run

macOS — double-click start.command, or run from terminal:

./start.command

Windows — double-click start.bat, or from a terminal:

start.bat

Both scripts install dependencies (if needed), check for your API key, start the dev server, and open your browser at http://localhost:3000.

You can also start manually:

npm run dev

---

Usage

1. Add reference images (optional) — drag and drop, click to browse, paste from clipboard, or enter a URL. Up to 10 images, max 10 MB each. Images are compressed client-side before sending.
2. Describe your concept (optional) — fill in any combination of Storyline, Subject, Environment, and Mood. All fields are optional; you can use images alone, text alone, or both.
3. Choose a prompt count — 3, 4, 5, or 6 prompts.
4. Generate — the pipeline runs server-side. With images, expect ~15–30 seconds for the vision step plus generation. Text-only is faster.
5. Revise — hover any result card, click Revise, describe what should change, and Apply. Only that card updates.

---

Stack

Layer	Tech
Framework	Next.js 16 (App Router)
Language	TypeScript 5
Styling	Tailwind CSS 4
AI Gateway	OpenRouter
Vision model	google/gemini-2.5-flash
Text model	minimax/minimax-m2.5

No database. No auth. No extra npm packages beyond the Next.js scaffold.

---

Project structure

src/
  app/
    page.tsx                  # Main page — state, generate + update handlers
    layout.tsx                # Root layout
    globals.css               # Global styles
    api/
      generate/route.ts       # Two-step pipeline: Gemini vision → MiniMax generation
      revise/route.ts         # Single-card revision: MiniMax rewrite
  components/
    ApiKeyInput.tsx           # Optional client-side API key input
    ImageUploader.tsx         # Drag-drop, paste, URL input
    InputForm.tsx             # Storyline / Subject / Environment / Mood fields
    PromptList.tsx            # Revision orchestration, visual style panel
    PromptCard.tsx            # Prompt display, copy, inline revise UI
  lib/
    openrouter.ts             # Typed fetch wrapper + JSON parser
    system-prompt.ts          # All prompt builders and shared types
    image-utils.ts            # Canvas resize/compress, clipboard, URL validation

---

Environment variables

Variable	Required	Description
OPENROUTER_API_KEY	Yes	Your OpenRouter API key — stored server-side only

---

License

MIT