Subtitle Converter

How it works

Supported SRT input has a positive numeric counter, a start/end line and one or more nonempty text lines per cue. Timestamps use hours:minutes:seconds,milliseconds. A blank line separates cues. Counter spelling is not treated as a timing instruction: SRT → WebVTT uses cue order, and converting WebVTT back generates new counters starting at 1. Cue start times must be non-decreasing, and every end must be later than its own start. Overlapping cues are retained and reported because simultaneous speakers or descriptions may intentionally overlap.

WebVTT begins with WEBVTT and a blank separator. Its timestamp decimal separator is a period, and a zero-hour timestamp may omit the hours. The output always writes hours explicitly. Converting WebVTT → SRT accepts a bare header and cues with no identifiers, NOTE blocks or settings. STYLE and REGION blocks, extra header metadata and region settings are refused by the parser. Other cue identifiers, a header comment or NOTE blocks stop conversion because those would otherwise disappear. A text-only converter is not a substitute for a full subtitle authoring system.

Conversion accepts plain cue text rather than inline markup. Angle-bracket formatting, speaker tags and embedded timestamps stop the operation. An ampersand in plain SRT text is escaped as & for WebVTT. The reverse direction decodes the six WebVTT named character references: amp, lt, gt, nbsp, lrm and rlm. Decoded angle brackets are refused because a receiving SRT player might interpret them as markup. Unknown character references are not guessed. This deliberately narrow behavior avoids pretending that every SRT player's informal HTML conventions mean the same thing.

UTF-8 files may have one initial byte-order mark; it is removed with a notice. Invalid UTF-8 bytes are rejected rather than replaced. LF, CRLF and CR source line endings are accepted, while output is UTF-8 with LF. Line-ending normalization and regenerated SRT counters mean the result is not a byte-identical archival copy, even when the visible plain text and timings are retained. Leading/trailing spaces inside cue payload lines remain data. Blank lines separate cues, so paragraph breaks inside one cue are outside this supported subset.

The bounded input is 2 MiB, 60,000 physical lines and 10,000 cues. Timestamp hours can extend to 9,999. A timestamp field must use valid minute and second ranges and exactly three fractional digits. Bad timing, missing counters, empty cues or a missing blank separator stops the full operation; there is no partial file that looks complete. The first 20 cues appear in the preview and every accepted cue appears in the download.

Caption text and file parsing remain in browser memory. The tool does not open a video, inspect audio or contact a subtitle provider. Editing the format or text removes the previous result, and a late file read cannot overwrite a newer edit. Clear removes the current widget text; it cannot erase files already downloaded, copied elsewhere or retained by your device. Review the exported file in the actual player, including language, cue timing and any intended overlap. Conversion alone does not certify accessibility or the correctness of the transcript.

Common use cases

• Prepare an existing plain SRT caption file for an HTML video text track
• Convert a plain WebVTT track for an application that requires SRT
• Spot unsupported metadata before a conversion would discard it
• Keep an unuploaded working copy of private interview or training captions

Frequently asked questions

Standards and limitations

• W3C: WebVTT file, cues and character references
• Library of Congress: SubRip structure and interoperability limitations