New Features

• AI Photo Styles — Define your own AI looks for customer photos, such as a cartoon or a line portrait sketch, under Design Assets › AI Photo Styles: give each style a name, thumbnail, prompt and image model, and try the prompt on a sample or uploaded photo before you save. Customers apply a style from the new AI Filters tool in the customizer; your prompts stay on your server.
• AI Photo Style on upload — The new print profile option "AI Photo Style on Upload" (Designs & Uploads › Advanced) applies one of your styles to every photo a customer uploads. The AI Photo Styles screen lists the print profiles that use each style.
• Manage categories from the product list — The Products page shows each product's categories in a new Categories column where you assign and remove them per row. Create, rename and delete categories in the "Manage Categories" modal and filter the list by category from the toolbar.
• Sortable print profiles — Sort the Print Profiles list by name or creation date from the column headers; your choice is remembered.

Improvements

• Image model and images per generation — The AI services settings let you choose the text-to-image model with its resolution in one dropdown that lists only valid combinations, each with its exact credits per image, and set how many images (1–4) one generation returns.

Bug Fixes

• Uploads failing with "Server response missing pdf_images" — A file larger than the server's upload limit produced an empty response and a technical error for the customer. The upload now answers with a clear message naming the server limit, and tells guests when an upload requires a login.
• Scanned PDFs rejected as suspicious — Harmless scanned PDFs were refused when their compressed image data matched a code pattern by chance. The security scan now flags only real code.
• Gravity Forms notification sent in English — The subject and first line of the Gravity Forms notification mail were never translated on German sites, and the Spanish and French versions still said "FPD". All three languages are corrected.
• AI Photo Styles without a name or prompt — Saving a style without a name could drop or delete it silently; it is now refused with an error. A style without a prompt is saved but never applied to uploads, and the print profile select marks it "(no prompt, not applied)".
• AI Photo Style select hard to use — Custom styles sat below the fold of the "AI Photo Style on Upload" menu with no visible scrollbar. The select is now searchable.
• Duplicate product categories — Creating or renaming a category to a title another category already has is now refused.
• Images per generation accepted any number — The "Images per generation" setting is rounded and limited to 1–4 on save.

Core Engine

• AI photo style applied to PDF pages — Pages of an uploaded PDF were sent through the upload slot's AI photo style. PDF pages are now placed unchanged.
• AI filters charged twice — Re-applying the same AI filter to the same image, or placing a stored upload again, ran a new AI request each time. Successful results are reused, so each image and filter is charged once.
• Processing animation for image filters — While a filter is applied, the image on the canvas turns softly out of focus and a scan line uncovers the result when it arrives, so customers see what is happening during longer requests. Reduced-motion settings are respected.
• Filtered images keep their size — An image keeps its displayed size and position when a filter returns different pixel dimensions.
• Newest uploads first — A newly uploaded image appears at the top of the upload grid, and stored uploads load newest-first. PDF pages keep their page order.
• Undo did not restore a filtered image — Undoing or redoing an image filter restored the filter state but left the other image on the canvas. The image itself now follows.
• Filters tool disappeared after filtering — The Filters tool stayed hidden when the server returned a result URL without a .jpg or .png extension, leaving no way back to the original.