Skip to main content

Changelog

September 29, 2026

What changed on Visxy on this date.

New

  • Publish your generations. Share a finished result with its prompt and inputs, choose which inputs others may reuse, and show your username or remain Anonymous. You can review and withdraw your publications from your profile or the Generator. See the Terms of Service and Privacy Policy for what becomes public.
  • Community in the Generator. Browse generations others have published, see their models, prompts and shared inputs, and open one with Recreate to make your own version. Your creations remain available in a separate view. Premium members can download published results.
  • Prompt library in the AI Generator. Browse and search image, video, effect and product prompts without leaving the Generator, then choose one to start from. Browsing does not use credits.
  • Point to pictures in your prompt. On supported models, type @ to refer to an attached picture by name. The Generator shows when a reference needs a picture before you generate.

Improved

  • Generate from any prompt page. Generate in Visxy is easier to find, with other tools one click away. If generation is temporarily unavailable, the Generator explains that before you start.
  • Motion Control in the AI Generator. Models that need a clip and a picture — Motion Control and the like — are in the AI Generator picker as well as on their tool page. The composer says which inputs are still missing, and Generate waits until both are attached.
  • Prompt length in the AI Generator. The selected model's prompt limit and a live count are shown before you generate. If your prompt is too long, you can shorten it without losing text or credits.
  • Writing @image1 before choosing its picture. In the AI Generator, the next picture you pick from your library now fills the tag you already wrote, instead of becoming @image2. The library shows which tag it will fill and lets you choose another one — or a new tag — and a tag whose picture you removed is only refilled if you choose it.
  • Switching model in the AI Generator. Your attached pictures and clips stay attached when you change model. The ones the new model uses the same way stay in use; the rest are dimmed and marked Off — not sent and not charged for — and come back into use when you switch back. A picture you attached as a reference is never turned into a video's first frame.
  • AI Generator help. The FAQ and writing guide now explain picture references, attachments, credit types and when your files expire.

Fixed

  • Pictures in an unsent AI Generator draft. The pictures and clips you had attached now come back with your prompt when you return to the page, including when a link opens a different model — there they wait, marked Off, until you switch back.
  • New attachments in a draft. A picture added while you start writing stays attached.
  • The AI Generator's attached pictures agree everywhere. The row of attachments and the picture library now show the same label for each picture — Start stays Start, End stays End — and count only the pictures that will be sent. Every picture that will be sent is always on the row; pictures kept aside for another model fold behind one +N Off button when the row is full, and stay visible even on a model that takes no pictures.
  • Using a picture again in the AI Generator. A picture you remove and pick again, or one kept aside for another model that you press in your library, is now used the way the model in front of you uses pictures — as its start frame on a video model, for example — instead of staying set aside for the model you first attached it to.
  • A reload in the AI Generator forgot what your @ tags meant. A tag whose picture you removed is no longer offered to the next picture you pick after reloading the page, and a picture you chose for a tag keeps that tag even if you leave before it has finished loading.
  • Using a set-aside picture for a tag. In the AI Generator, pressing a picture that is set aside for another model now fills the tag chosen at the top of your library, as any other picture would — and if your prompt already uses that picture's own tag, it keeps it and tells you.
  • Picking a removed picture again. In the AI Generator it gets its own @ tag back, instead of taking the tag your library suggests for a new picture.
  • Long prompts on PixVerse V6 and Wan 2.7. The AI Generator now tells you before you generate when a prompt is longer than these models accept — 5,000 characters on Wan 2.7, and 2,048 bytes on PixVerse V6, where accented letters and emoji count double or more — instead of letting the request fail.
  • Reusing an older generation. When its picture references need review, Review in composer restores the prompt and attachments so you can update them before generating again.

Policies

  • Your rights and contact details. Our legal pages now give clearer information about refunds, privacy choices, data retention, account deletion and how to contact us.