Subtitle & Transcript Tools That Never Leave Your Browser

Subtitle and caption files are just structured text. An SRT (SubRip) file lists sequential cues: a number, a timecode range in HH:MM:SS,mmm --> HH:MM:SS,mmm format using a comma before the milliseconds, and one or more lines of dialogue, separated by a blank line. A WebVTT (.vtt) file opens with a WEBVTT header, uses a period instead of a comma before milliseconds, and supports extra metadata SRT doesn't: cue settings like align:start position:10%, voice tags such as <v Speaker Name> for per-speaker captions, and styling tags like <i> or <c.yellow>. Plain .txt files have none of this, just lines of text, which is exactly what most people actually want to read, copy into a document, or feed into a summarizer.

Step 1: Load your file

Drag a .srt, .vtt, or .txt file onto the Studio above, click “Upload File,” or just paste the raw text directly into the left pane. SubToolDesk reads the file locally with the browser's FileReader API. It is parsed in memory and is never transmitted anywhere.

Step 2: Let auto-detection do its job

The badge above the editor identifies the format by scanning for a WEBVTT header, comma-style timecodes, or dot-style timecodes. If none of those patterns match, the file is treated as plain text and each non-empty line becomes its own segment, so speaker-label and tag cleanup still apply.

Step 3: Pick an output and tune the toggles

Choose Plain Text for a clean, readable transcript, or SRT/VTT to keep (or convert) the timed cue structure. Three independent toggles refine the result:

• Merge into paragraphs: stitches short, fragmented cues back into natural sentences and paragraphs, using punctuation to decide where one thought ends and the next begins.
• Strip speaker labels: removes leading identifiers like SPEAKER 1:, [John]:, or >> JANE: that captioning software adds automatically.
• Strip formatting tags: removes HTML-style tags (<i>, <v Name>) and ASS override codes ({\an8}) that otherwise leak into a plain-text export as visual noise.

Step 4: Copy or download

The right-hand pane updates on every keystroke and every toggle change. When it looks right, press Copy (or Cmd/Ctrl + C anywhere outside the text panes) to grab it from the clipboard, or Download to save a correctly-extensioned file, all without a single network request leaving your machine.

SRT vs. WebVTT: the technical differences that matter

The two formats look almost identical at a glance, which is exactly why conversions between them trip people up. SRT timecodes separate milliseconds with a comma (00:00:06,120); WebVTT uses a period (00:00:06.120). WebVTT requires a literal WEBVTT header line at the top of the file. SRT has no header at all. WebVTT also supports features SRT was never designed for: cue settings like align:start position:10% that control on-screen placement, <v Speaker Name> voice tags for per-speaker captioning, and dedicated NOTE, STYLE, and REGION blocks for comments and styling. SRT's cue index must be a strict, gap-free sequential integer; WebVTT's optional cue identifier can be any string, or omitted entirely. Getting any one of these details wrong is usually why a hand-edited subtitle file "almost works" in a player but silently fails to load captions in a browser.

Who actually uses tools like this

Video editors convert timed caption exports into a plain-text script for a quick read-through before a cut is locked. Podcasters and content teams turn an auto-generated transcript into a blog post or show-notes draft without re-typing anything. Customer success and sales teams pull a clean, speaker-free summary out of a Zoom or Teams recording for a CRM note. Web developers prepare .vtt files for the HTML5 <track> element to meet accessibility requirements. Researchers and journalists clean up interview transcripts for quoting without exposing the raw recording to a third-party transcription vendor a second time. Each of those use cases shares the same underlying requirement: the source file is often sensitive, pre-release, or simply none of a stranger's server's business.

A note on accuracy and limits

The paragraph-merging algorithm uses punctuation (periods, question marks, exclamation points, and the ellipsis convention subtitle authors use to mark a sentence continuing into the next cue) to decide where one paragraph ends and the next begins. It's deliberately conservative: when in doubt, it keeps a fragment attached to its neighbor rather than guessing at a break that isn't there. For files with unusual punctuation conventions (no terminal punctuation at all, for instance), toggling paragraph merging off and reviewing the one-cue-per-line output first is the fastest way to sanity-check the source before reformatting it further.