1 Add an API Key (many are FREE)

This tool uses AI to write your prompts, so it needs an API key. Click the AI Engine box in the sidebar to open AI Settings. Choose from 7 providers: Google Gemini (free tier - best to start), Groq (free, very fast), Mistral AI, OpenAI, OpenRouter (one key unlocks Gemini, Claude, GPT, LLaMA), Anthropic Claude (top-quality vision), xAI Grok (free starting credits). Pick a model, paste your key, press Add, then Save.

2 Pick Your Mode

Image - upload images and get prompts that recreate or vary them. Video - upload video clips; a key-frame is extracted automatically. Text Idea - just type an idea and get polished prompts, no upload needed.

3 Choose the Target AI Tool

Select where the prompt will be used - Midjourney, Stable Diffusion / SDXL, DALL-E, Leonardo, Flux, Firefly or video tools like Sora, Runway, Pika, Kling, Veo. Each has its own prompt syntax, and the output is formatted to match it (including --ar, --v style flags where relevant).

4 Customize the Style

Shape every prompt with: Art Style, Mood / Lighting, Aspect Ratio, Detail Level, Camera Angle / Shot, Color Palette, Creativity (precise → imaginative), Variations (how many prompts per file), and Output Language. Art Style, Mood and Aspect Ratio each have an Auto choice - pick it when you want the prompt to follow whatever the uploaded file already looks like instead of forcing a different style.

Prompt Match Mode (image & video modes) decides the goal: Exact Match writes prompts that reproduce your file as closely as possible, while Creative Variation keeps the subject and style but changes pose, angle or background so you get a new picture. In Exact Match the tool automatically softens any style override that would fight the original - accuracy wins.

Toggles let you include camera / lens details, platform tags, and negative prompts. Two free-text boxes give you full control: Custom Instructions (added to every request - e.g. "avoid people, always autumn colors") and Custom Negative Words. All settings save automatically.

5 Selling on Stock Sites

If you plan to sell the generated images on Adobe Stock, Shutterstock, Freepik and similar sites, keep Stock-Ready Prompts switched on. It adds the rules that decide whether a file is accepted or rejected: no brands or logos, no readable text (AI writes gibberish letters - the number one rejection reason), no recognizable or celebrity faces, no copyrighted characters or artworks, correct anatomy (five fingers, natural eyes), and sharp artifact-free quality.

Four options tune it for what buyers actually pay for: Copy Space leaves an empty area for headlines (designers need this), Background gives you isolated-on-white or cut-out-ready PNG looks that sell strongly, Commercial Theme aims the subject at high-demand categories like business, healthcare or technology, and People in Shot lets you avoid people entirely or keep faces unidentifiable so no model release is needed.

Text in Image has three settings, because text needs opposite handling depending on the job:

No text (default, safest) - the prompt actively forbids lettering and signage, since garbled AI text is the top rejection reason for normal stock photos and illustrations.
Keep the text from my image - when your uploaded file contains wording, the AI reads it and repeats it in the prompt character for character, so the new image keeps the same text. Best combined with Exact Match.
Text design - for t-shirt, typography, quote and poster designs, where the lettering is the product. Type the exact words, pick a lettering style and layout, and the prompt demands perfect spelling, clean print-ready composition and an isolated background for cut-out. It also blocks trademarked slogans, lyrics and movie quotes, which stock sites reject.

6 Generate

Drag & drop files (or press Ctrl+V to paste an image), or type an idea in Text mode and press Generate. Use Pause, Resume and Stop while a batch runs. Failed items show the reason and can be re-run with Retry.

7 Use Your Prompts

Each result card lets you copy a single prompt, copy all variations, or Regenerate that item for fresh ideas.

Four export choices sit in the toolbar, and every one keeps prompts clearly separated instead of running them together:

CSV - one numbered row per prompt for spreadsheets, with a blank row between files.
TXT - a plain file with a header, boxed titles per source and blank lines around every prompt.
HTML - the most useful one: it downloads a small web page you can open any time, even offline. Every prompt sits in its own card with a Copy button, plus a search box to filter prompts and a Copy all button. It follows your system dark or light theme.
Copy All - copies everything to the clipboard with clear separators between each prompt.

Every prompt carries two numbers so nothing gets lost in a big batch: a running number across the whole export (#7 of 20) and its position inside its own source (Prompt 3 of 4). The same pair is shown on screen, and the header tells you how many sources and how many prompts you have in total.

Privacy

Everything runs in your own browser. Your API keys and settings stay on your device - nothing is uploaded to our server. Files go only to the AI provider you chose, using your own key.