This is a browser page, so it is deliberately limited. It reads what a browser can decode, which
means MP4, MOV and WebM behave well while AVCHD, MTS, MXF and older camera formats will often show up
as not readable here. It also forgets everything the moment you close the tab, and it uses the
file date on disk, which is usually the date a clip was copied rather than the date it was shot.
Copies are spotted by matching file size and filename together, so a copy that was renamed beyond
recognition slips through here.
Those three limits are exactly what the built version fixes:
Every format, properly read. A desktop tool using ffprobe reads length,
resolution, frame rate, codec and the real embedded creation date out of every file, including the
formats this page has to skip.
The index survives unplugging the drive. Each drive is scanned once and remembered,
so you can search all of them at the same time with none of them connected, and see which drive a clip
lives on before you go looking for it.
Duplicates found by content, not by name. Matching on a fingerprint of the file
itself catches the same clip saved under three different names in three different years, which is the
version of this problem that actually wastes your afternoon.
Published lesson to source clip. The finished video in your library points back at
the raw clips it was cut from, so deciding to remake something stops being the start of a search.
The keep / remake / needs-your-eyes queue. Your transcripts plus a frame or two per
video, sorted against your criteria rather than mine. Every call stays yours. You just make a
hundred of them by looking at twenty.
The flags on the cards above are rough demo heuristics, not judgements about your teaching: a clip is
flagged when another file matches it, or when it is below 720p. Nothing here is analysing whether a
lesson is any good.