One rough idea in. A batch of finished prompts out.

What can you start from?

With no persona picked, a prompt written from a photo invents its own subject and never copies the person in it.

• An idea: one line like "coffee shop mornings" becomes as many prompts as you ask for
• Photos: paste, drop or browse, and get one prompt per photo that recreates its scene
• Prompts you have: improve one, make three variations, or ask for more like the ones that worked

How does it keep a persona consistent?

Every shot with a persona or product follows a master prompt. A persona, a product or a combination of them can have its own; otherwise the shot follows the default for its kind (a persona alone, a product alone, a persona with a product, and so on). The built-in default for people describes the scene, outfit, pose and light, and leaves the face, hair and skin to the reference photos, with a casual phone-camera look. The one for products keeps the product exactly as photographed and writes a clean, professional product shot.

The workspace owner can rewrite any of them, and the ones used for writing from a photo and for Improve, Vary and More like these. Whatever they say, every request also tells the model never to name or depict a real, identifiable person.

Which AI model writes the prompts?

Any model on OpenRouter, on the workspace's own key, picked by the owner in Settings. Personifex sets no daily limit: your OpenRouter balance is the only one.

Where do the prompts go?

Drafts stream into a preview, and nothing joins the queue until you press Add, which you can undo. Then they sit in the persona's queue in the Prompts tab, shared with the team, next to a history of what ran (with the reason when something failed) and every prompt sorted by the rating its image got. Search, reorder, export to a text file, or copy prompts into another persona's queue.