Visemory searches local video by dialogue lines and scene descriptions

On October 3, 2026, the developer of Visemory introduced the tool in a post on V2EX, describing it as a way to search local video, audio, and image files using natural language. The post frames the problem plainly: people remember that a certain interview contained a particular line, or that they once shot a usable scene, but they cannot recall which file holds it or at what minute it appears, so they end up opening videos one by one and dragging the progress bar.

Visemory is meant to solve that by searching local media by content and locating the exact line of dialogue or the specific shot. Users add local folders holding video, audio, and images, and once a first-time processing pass finishes, the material becomes searchable without migrating or copying the original files. Searches can target dialogue and subtitles, either by typing a remembered sentence to jump to the matching clip or by topic and meaning when only the gist is recalled. They can also target visual content, with users describing the shot they want in natural language, such as "mountains and scenery". A third option covers text that appears on screen, such as text in course slides, presentations, or screen recordings. Cross-language search lets users query in their own language for material in another, so a search for a greeting like "good morning" can surface clips containing "Hello".

According to the post, results point to the relevant time codes and support preview and export. All media processing is done locally by default, which the developer presents as protecting data privacy. The post also shows a result for the query "mountains and scenery", with the matching frames displayed. The developer says he is looking for 10 people who frequently dig through local footage to try the tool. No further milestones or version details have been reported.