Bulk find and replace in subtitles
Fix homophone typos and inconsistent terminology across a whole file. Preview every change before you commit — only the subtitle text is touched; numbering, timing and inline tags stay as they are.
Accepts .srt / .vtt up to 5 MB.
The file is never uploaded — everything runs locally in your browser
Want to handle the video at the same time?
The full version takes the video itself: it transcribes the speech, translates it into 20 languages, and — after you fix terminology in one place — burns the bilingual subtitles into a finished file.
Sign up free and get trial creditsFAQ
Why can’t I just use a global replace, the way I would in a text editor?
Because an SRT is not plain text — it has three parts: the numbering, the timing, and the text. A global replace in Notepad changes the numbering and the timecodes too. If your search term happens to sit inside a timecode (say you replace “00”), you destroy the whole file’s timing, and the player reports no error — the subtitles just end up out of sync. This tool only rewrites the text lines; the numbering and the timing are carried over unchanged.
Why does “Find and preview” not change anything, and why do I have to click apply afterwards?
One bad batch edit is worse than a dozen small ones. The preview lists every change one by one (old line → new line, with the match highlighted), so you can glance over it, confirm the rule does what you meant, and only then click Apply. That step matters most in regex mode — a single wrong pattern can wipe out every subtitle in a fraction of a second.
What does “Protect inline tags” do?
SRT files sometimes carry inline tags such as `<i>` (italic) or `<font>`. By default the search skips the contents of those tags and applies only to the text around them, so a replacement cannot turn `<i>` into something else and break the tag. If your target really is inside a tag, turn the switch off.
What else is worth watching out for?
1. “Whole word” has no effect on Chinese — Chinese has no word boundaries, so the option only matters for Latin letters and digits. 2. In regex mode you can reference capture groups with $1 and $2; in literal mode the $ is treated as an ordinary character (common in passwords and amounts). 3. This tool does no “smart correction” and will not guess which character is wrong — a wrong fix is hard to spot by eye, so it only runs the exact rules you give it.