🎬 captionkit

Convert and re-sync subtitles — SRT ⇄ WebVTT, shift, fix drift — locally.

🌐 Try the free web app →  ·  drop a .srt or .vtt, fix the sync, download. Nothing uploaded.

---

You exported captions and the player won't take them (it wants WebVTT, you have SRT). Or the subtitles play 2 seconds late. Or they slowly drift out of sync by the end of the video. Fixing this by hand means editing dozens of 00:01:23,456 timestamps without breaking the format — and one stray comma or missing millisecond makes the whole file invalid.

captionkit does it correctly: convert SRT ⇄ WebVTT, shift every cue, scale/resync to fix drift, and fix overlaps — with exact millisecond timecode math, zero dependencies, and 100% locally.

📸 Screenshot / demo GIF: ./web/screenshot.png — record the live app dropping an out-of-sync SRT and nudging it into place.

Why it exists

• AI can't reliably do this. Re-timing every cue and emitting spec-valid SRT/WebVTT (comma vs dot, WEBVTT header, numbering rules, CRLF) is exact, fiddly format work — a chatbot will quietly corrupt a timestamp. It's a job for a small, tested, deterministic tool.
• Privacy & friction. Online subtitle converters make you upload your file and wait. captionkit runs on your machine, instantly.
• The drift fix is the "aha". resync linearly maps the first and last cue to where they should be — repairing subtitles that slip out of sync as the video goes on. Most tools only do a flat shift.

Who it's for

Video creators & YouTubers, marketers (captioned social video), educators, accessibility teams, translators, and developers building media tooling who want a tiny subtitle library.

Install

No install — just open the web app.

For the library:

npm install captionkit

Zero dependencies. ESM + CJS + TypeScript types. Runs in the browser, Node, Deno and Bun.

Usage

import { parse, convert, shift, resync, toVTT } from "captionkit";

// Convert SRT → WebVTT (auto-detects the input)
convert(srtText, "vtt");

// Parse, then re-time
const cues = parse(srtText);
shift(cues, 2500);                 // everything 2.5s later
shift(cues, -1000);                // …or earlier (clamped at 0)

// Fix drift: map the first cue to 1.0s and the last to 10:00.0
resync(cues, { firstStart: 1000, lastStart: 600000 });

toVTT(cues); // serialize back out

More re-timing

import { scale, fixOverlaps, totalDuration } from "captionkit";

scale(cues, 25 / 23.976);   // framerate conversion
fixOverlaps(cues, 40);      // no two cues overlap (40ms min gap)
totalDuration(cues);        // total on-screen time (ms)

API

Function	Description
parse(text)	Parse SRT or WebVTT → Cue[] ({ index, start, end, text }, ms).
toSRT(cues) / toVTT(cues) / convert(text, to)	Serialize / convert.
detectFormat(text)	"srt" or "vtt".
shift(cues, ms)	Offset all cues.
scale(cues, factor)	Multiply all timestamps.
resync(cues, { firstStart, lastStart })	Linear drift correction.
fixOverlaps(cues, minGap?)	Remove overlaps.
renumber / totalDuration	Helpers.
parseTimecode / formatSRT / formatVTT	Timecode ⇄ ms.

FAQ

Is my subtitle file uploaded anywhere? No. Everything runs on your device — no server, no telemetry, works offline.

What formats are supported? SRT and WebVTT today. ASS/SSA and SBV are on the roadmap — open an issue.

What's the difference between shift, scale and resync? Shift adds a fixed offset (captions are uniformly early/late). Scale multiplies timestamps (framerate mismatch). Resync pins the first and last cue to target times and interpolates the rest — the fix for gradual drift.

Will it keep my line breaks and styling? Multi-line cue text is preserved. Inline WebVTT tags pass through as text; full ASS styling is not converted (yet).

Can I use it in a build or batch script? Yes — the library works in Node; map the functions over your files.

Contributing

Contributions welcome! See CONTRIBUTING.md and the Code of Conduct.

git clone https://github.com/didrod205/captionkit.git
cd captionkit
npm install
npm test          # run the suite
npm run dev       # run the web app locally

💖 Sponsor

captionkit is free, MIT-licensed, and built in spare time. If it saved you from hand-editing timestamps, please consider supporting it:

• ⭐ Star this repo — free, and it genuinely helps others find it.
• 🍋 Sponsor via Lemon Squeezy — one-time or recurring support.

Where your support goes: more formats (ASS/SSA, SBV, SAMI), split/merge, characters-per-second warnings, an auto-shift-by-waveform helper, a CLI, keeping the free web app online, and fast issue responses.

License

MIT © captionkit contributors