Auto-Caption Cleaner: Turn Broken Subtitles Into Readable Text
Paste a YouTube or Bilibili auto-caption file (SRT or VTT). Lumi removes the rolling duplicates, re-merges fragments into complete sentences, restores punctuation, and gives you a clean .srt or .txt transcript. Everything runs in your browser, so no file leaves your machine.
What this tool repairs
| What it fixes | Details |
|---|---|
| Rolling duplicates | YouTube rollup captions repeat the previous line in every cue. Duplicates are collapsed before anything else runs. |
| Fragments cut mid-clause | Cues are regrouped by pause length, duration and a character budget, so a sentence is no longer split at an arbitrary word. |
| Missing punctuation | Sentence-final punctuation is restored; Chinese output uses full-width marks instead of Latin ones. |
| CJK spacing | Spaces between Chinese characters are removed while Latin words keep their separators. |
| Inline markup | Word-level timing tags, <c> wrappers and HTML entities are stripped from the output. |
| Malformed blocks | SRT and WebVTT are detected automatically, and an unparseable block is skipped instead of failing the whole file. |
Questions
- Why do YouTube auto-generated subtitles look broken?
- Auto-captions are produced word by word as the video plays, so each cue is a fixed-width chunk of speech rather than a sentence. The WebVTT version additionally repeats the previous line in every cue to build the rolling effect, which is why a raw download shows duplicated lines, clauses cut in half and no punctuation at all.
- Does this tool send my subtitle file to a server?
- No. Parsing, duplicate collapsing and re-segmentation are deterministic rules that run in your browser. There is no upload endpoint, no account and no server-side step on this page, so the file stays on your machine even if you later paste client-confidential material into it.
- Can I use the cleaned file in a video editor?
- Yes. The .srt export keeps valid numbering and timestamps, so it imports into Premiere, DaVinci Resolve, Final Cut, VLC or back into YouTube Studio as a corrected caption track. The .txt export is for scripts, notes and feeding a language model.
- How is this different from the Lumi extension?
- This page repairs a subtitle file you already have. The extension does the same re-segmentation live beside the player, then adds bilingual translation, timestamped chapter summaries, grounded Q&A and a searchable library, without exporting files first.
The extension runs the same repair live beside the player, then adds bilingual translation, chapter summaries and Q&A.