What is new in Fast Video Cataloger 10.3
Fast Video Cataloger 10.3 is now available for download.
You can now connect an AI assistant to your catalog. The installer includes a connector built on the Model Context Protocol, the open standard assistants such as Claude use to reach outside systems. Point an assistant at your Fast Video Cataloger server and ask for footage in your own words: it searches your videos, scenes, transcripts and cast, looks at the actual scene thumbnails to judge what is in a shot, and can tag videos and gather clips into bins for you. Everything runs through your own server, so the catalog and the video files stay on your machines. Setup is described under Connect an AI Assistant in the server guide.
You stay in control of what an assistant is allowed to do. Connecting one means giving it a key, created in the User Management window, and the key decides its access: give out a read only key and the assistant can search and look but never change anything in your catalog, and you can revoke a key at any moment. Videos already in the catalog that were never indexed can now be indexed through the server as well, so a script or an assistant can work through a backlog and give it thumbnails and keywords, applying whatever the server is set up for, including transcription and the AI features. For scripting, the server API has grown in several places as well; the details are in the version history.
Trained Tags let you teach the AI keywords of your own. Select around ten thumbnails that show the thing you mean, right click and choose Teach trained tag, and Fast Video Cataloger can then find that subject across the catalog and keyword it, with no dataset to annotate and no training run to wait for. A tag can match whole frames, for scenes and settings like a particular studio, or apply to a detected object so you can tell apart kinds of the same thing. Matches become purple keywords, and a wrong match is corrected by right clicking it and choosing Remove and teach as, which itself teaches the right tag. The new Trained Tags window on the start page manages your tags and can test them on the current selection before a real run.
AI image tagging can now run while videos are indexed, instead of only as a batch afterwards, so new material arrives in the catalog already tagged. Tagging also now adapts to your own footage: the program learns what is normal for each catalog, so a library made up of one particular kind of video, such as screen recordings or interviews, gets more selective and more useful tags with nothing to set up.
Day to day work gets a few improvements as well. The Search window now has List and Tree checkboxes above the keywords, so you can switch between the flat keyword list and the keyword tree without going to Preferences. The Keyword Manager can filter its list by group, remembers its column layout, and no longer opens with the first keyword selected. The colour square next to a keyword in the keyword tree now shows a tick when the keyword is applied to the selected videos.
The release also brings a long list of improvements across search, keywording, the File Organizer, the server and the installer. The full list of changes is in the version history in the user guide.
Download the latest version from our download page. If you already run Fast Video Cataloger 10, the built-in update check will offer you the new version.