Words to SRT
Make SRT or WebVTT captions from word timestamps
Drop a transcript with word times (OpenAI verbose_json, or a Sucia Studio transcript), choose how many words a caption holds, and download SRT or WebVTT.
Free. Works in any browser, on a phone or a computer. What you put here stays on your device.
About Words to SRT
What it reads
- OpenAI-style verbose_json: a JSON object with a words list, each word with word, start and end in seconds (what the transcription API returns with word timestamps on). The same words inside segments are read too.
- A Sucia Studio transcript file: segments with words, each word with s, e and w in seconds; one chunk file or a list of them.
How the captions are cut
- At most the number of words you choose a cue, 8 to start, and at most the seconds you choose, 3.5 to start.
- A new cue after a pause of 0.7 seconds or at the end of a sentence; every cue stays up for at least 0.2 seconds.
- The same rules Sucia Studio uses for its captions, from the same code.
How to use it
- Drop the JSON file, or paste it. It is read on this device.
- Set the words and seconds a cue; the preview changes as you do.
- Download the SRT or the WebVTT and upload it in YouTube Studio (Subtitles), which takes both in plain UTF-8.
Rules as read on October 9, 2026 from YouTube Help: Supported subtitle and closed caption files.
Questions
- Where do I get word timestamps?
- From OpenAI's transcription API with word timestamps on (verbose_json), from Whisper tools that write the same shape, or from Sucia Studio, which makes a transcript on your computer.
- Why eight words a cue?
- It is what reads well on a phone: two short lines, gone before the next thought. Set fewer for Shorts, more for a slow talk.
- Can it fix the words?
- No. Fix the text in the JSON first, or open the SRT it makes in Caption Check and fix the lines there.
- Is my transcript uploaded?
- No. It is read and converted in your browser.