Extract hardcoded subtitles
Burned-in subtitles in, editable subtitle file out. Box the text on the picture, and every line it finds comes back with its own timing — SRT, VTT, ASS, plain text. Your video is read on this device; nothing is uploaded.
Two ways in: text burned into the picture (hardsubs — box it and read it), or spoken audio (a local speech model writes the draft for you). Either way nothing is uploaded.
subextractor folder run
python3 -m http.server 8000 and open http://127.0.0.1:8000/extract.html.
Drag a box over the subtitles. Box just the subtitle strip — leave faces, logos and busy background out. Fine-tune it with the handles; ‹ › steps one frame at a time.
Seek the video to a moment with a subtitle, then try it.
Reads the whole video on this device. First run loads the recognition models once (about 30 MB, kept by your browser).
Click a timestamp (or the ▶ button) to jump the video there — the line under the playhead stays highlighted. Each line also has add a line after, merge with the next and delete.
Nothing matches that search.