You cannot remember a travel video’s filename, but a red umbrella comes to mind. Or perhaps a sign, or something someone said. These invented examples point to three different kinds of search clue.
One video, different clues
SCM for macOS documents visual scene search, OCR for visible text and literal searches of transcribed dialogue. Its scene pipeline indexes each segment’s midpoint frame, rather than every frame.
Languages and scope
Korean OCR does not establish Korean dialogue support: the documented transcription models are English-only tiny.en and base.en. We have not tested retrieval quality or audited privacy.
V’s view
V’s view. Before rewriting a query, ask whether your memory is a shape, a written word or a sound. If the umbrella flashed past briefly, the next question is whether a representative frame captured it.