What changes when converting SRT, WebVTT, SBV, and ASS subtitles?
Subtitle formats differ in file headers, millisecond separators, and styling syntax, so any SRT to VTT, SBV, or ASS conversion normalizes every timestamp into integer milliseconds first. Once parsed into uniform cue records, the worker applies timing math, text filters, and sorting before writing .vtt, .srt, .sbv, or .txt output.
| Format | Header & cue index | Timestamp delimiter | Extra metadata & styling |
|---|---|---|---|
WebVTT (.vtt) |
WEBVTT header, optional cue IDs |
Period (00:00:01.200) |
Skips NOTE/STYLE/REGION; keeps cue alignment settings |
SubRip (.srt) |
Sequential integers 1, 2, 3 |
Comma (00:00:01,200) |
Supports HTML tags such as <i> and <b> |
YouTube (.sbv) |
No header or index numbers | Period (0:00:01.200,0:00:04.500) |
Comma-separated start and end with trimmed leading hour zero |
ASS / SSA (.ass) |
[Script Info] and [Events] Format: |
Centiseconds (0:00:01.45) |
Input-only; decodes \N/\h and {\...} overrides |
Plain text (.txt) |
None | Removed completely | Output-only; joins cleaned dialogue lines chronologically |
Timestamp delimiters, headers, and ASS centisecond math
SubRip expects HH:MM:SS,mmm with a comma before milliseconds, whereas HTML5 players following the W3C WebVTT Video Text Tracks Specification require a WEBVTT magic header and period separators such as 00:00:01.200. WebVTT parsing also ignores NOTE, STYLE, and REGION blocks cleanly without injecting comment text into dialogue captions.
ASS and SSA scripts store timings inside Dialogue: lines mapped by the [Events] Format: column list (Layer, Start, End, Style, Name, MarginL, MarginR, MarginV, Effect, Text). Two-digit ASS centiseconds such as .45 multiply cleanly to 450 milliseconds, while \N and \n turn into line breaks and \h expands to a normal space.
Preserving WebVTT cue settings vs renumbering SRT blocks
When converting to .vtt, trailing cue positioning parameters after the --> line such as alignment or percentage placement stay attached to each output cue. Converting to .srt omits WebVTT settings and renumbers every surviving block sequentially from 1 so empty cues dropped during cleanup never leave broken gaps.
Dropping any .vtt file flips Output format from WebVTT to SubRip automatically when needed, saving an extra manual click. When caption files represent Zoom or Teams calls rather than video tracks, use the transcript workflow to clean speaker-labeled meeting transcripts into speaker-merged notes.
Fixing subtitle drift with millisecond shifts and FPS scaling
Subtitles fall out of sync for two distinct reasons: a constant trim at the head of a video, or a playback speed difference between film (23.976 fps) and PAL broadcast (25 fps). Choosing between a fixed offset and frame-rate scaling depends on whether the sync error stays flat or grows worse toward the credits.
Constant offset nudging and zero-floor clamping
A constant offset happens when a distributor inserts a 2 seconds studio bumper or trims an intro without shifting captions. Type an exact positive or negative integer under Shift timing (milliseconds) or tap -1000, -500, +500, and +1000 buttons (500 ms or 1000 ms nudges) to preview recalculated cue boundaries immediately.
Every shifted start time clamps at 0 ms (00:00:00.000) so negative offsets never emit invalid negative hours into players following Matroska Subtitles and SubRip Codec Mapping conventions. Each cue end timestamp enforces Math.max(startMs + 50, ...) so ultra-short captions remain visible for at least 50 ms.
Correcting progressive drift with FPS speed ratios
When captions match at minute zero and lag by several seconds after an hour (a 4.27% speed gap between 23.976 and 25 fps), the subtitle file and video use different frame rates. Under FPS speed correction, select presets covering 23.976 fps -> 25 fps, 25 fps -> 23.976 fps, 23.976 fps -> 24 fps, or 29.97 fps -> 30 fps, or pick Custom FPS ratio at 0.001 precision.
For every cue, the worker computes Math.max(0, Math.round(ms * (fpsFrom / fpsTo) + shiftMs)), combining multiplicative speed correction and constant offset shifting in one pass. After adjusting timing across long scripts, run dialogue lines through our dialogue word count and reading pace checker to check overall reading load.
Stripping ASS tags, SDH cues, and overlapping cue timings
Auto-generated captions, fan-subbed ASS tracks, and broadcast SDH files frequently contain curly-brace styling tags, inline karaoke timestamps, or sound effects that break plain players. Accessibility guidelines under the W3C WCAG 2.2 Captions (Prerecorded) Standard require non-speech descriptions for deaf and hard-of-hearing audiences, though plain-text readers or dubbing scripts often want clean spoken lines only.
Removing styling markup and SDH brackets safely
Enabling Strip HTML & ASS styling tags removes ASS {\...} override blocks, WebVTT inline <HH:MM:SS.mmm> timestamps, and HTML tags such as <i>, <b>, <u>, or <font>. Enabling Remove [music] & SDH descriptions strips bracketed tags ([...]), common sound-effect parentheses (music, applause, sigh, laugh, sfx, noise, cheering, cough), musical note spans (♪...♪), and uppercase SPEAKER: prefixes.
Any cue whose dialogue becomes completely empty after tag and SDH filtering is dropped automatically and counted under removedEmpty instead of leaving blank timed cards onscreen. Repeated whitespace collapses cleanly per line while preserving deliberate multi-line subtitle breaks.
Sorting out-of-order cues and resolving overlaps
Edited tracks sometimes contain cues out of chronological sequence or overlapping intervals where cue 2 lingers past cue 3’s entrance. All cues sort by startMs before export, and keeping Fix overlapping cue timings checked clamps any colliding end time to next.startMs - 1 (respecting the 50 ms cue floor) while reporting how many overlaps were fixed.
Tip: Click Load sample SRT (
product-walkthrough.srt) to watch<i>italic tags,[gentle music],(applause), and the built-in overlap between cues2and3clean up automatically in one pass.
Before committing revised subtitle tracks into a repository or localization pipeline, inspect the modified text output using timed cue edits ready for line-by-line diff review against your original caption file.
Browser worker privacy, MIME exports, and companion formats
Subtitle drafts frequently belong to unreleased films, internal product launches, or confidential training courses that should stay off external web servers. Every BOM strip (\uFEFF), CRLF normalization, regex cleanup, and timing calculation runs locally via runEverydayTask inside a browser Web Worker. No cue text or speaker timing leaves your machine.
Downloads carry format-accurate MIME headers: text/vtt;charset=utf-8 for .vtt, application/x-subrip;charset=utf-8 for .srt, and text/plain;charset=utf-8 for .sbv and .txt. The live status bar updates after each adjustment, displaying detected input format, target format, cue count, total duration (HH:MM:SS.mmm), and fixed overlaps before you click Copy output or Download.
Once saved, ResultPanel confirms the output file name and byte count locally without retaining any transcript text in remote storage. Output file names preserve your original upload stem paired with the selected extension. Teams converting other fixed-format local documents alongside media caption workflows can also open Windows print-spool containers offline with XPS to PDF.