How to Compare Two Subtitle Files (SRT/VTT)
You have two versions of a caption file — an original and an edit, a human transcript and a machine one, or two translations — and you need to know exactly what changed in the wording and the timing. Scrolling two SRT files side by side is slow and error-prone. This guide shows you how to compare subtitle files objectively, catching both text and timing differences.
Two kinds of difference
Subtitles have two dimensions that can drift independently:
- Text — the actual words. An edit might fix a typo, change a translation, or drop a line entirely.
- Timing — when each cue appears and disappears. A re-sync, a framerate conversion, or a trimmed intro can shift every timestamp.
A good comparison reports both, separately, so you know whether the words changed, the timing changed, or both.
The wording score: Word Error Rate
Word Error Rate (WER) is the standard metric from speech recognition for how much two transcripts differ. Rather than a raw character diff, it captures meaningful edits — words substituted, added, or dropped — as a single percentage. WER of 0 means identical wording; a low WER means a handful of small edits; a high WER means a genuinely different script or translation.
DiffALL computes WER for you and shows it as a headline number, so "how different is the wording" becomes one figure instead of a manual line-by-line read-through.
Doing this by hand is the hard part. Calculating WER and aligning timestamps across hundreds of cues means writing a script, handling encoding quirks, and parsing two subtitle formats correctly. DiffALL handles all of it from a drag-and-drop.
The timing report
DiffALL lines up the cues and reports how far each one shifted. A constant offset across every cue means a simple re-sync (the whole file is early or late). A drift that grows over time points to a framerate mismatch — the classic 23.976 vs 25 fps problem — where subtitles start in sync and slide further off as the video plays.
Step by step
- Open DiffALL — no install or account needed for your first comparison.
- Drop both subtitle files (SRT or VTT) into the upload zones.
- Run the comparison to get a WER score for the text and a timing drift report.
- Review the line-by-line diff to see exactly which cues changed in wording or timing.
Reading the results
| What you see | What it means |
|---|---|
| WER = 0, timing identical | Files are effectively the same. |
| Low WER, timing identical | Minor text edits — typos, punctuation, a reworded line. |
| WER = 0, constant timing offset | A clean re-sync — same words, shifted in time. |
| WER = 0, growing timing drift | Framerate mismatch (e.g. 23.976 vs 25 fps). |
| High WER | A different translation, transcript, or heavily edited script. |
Common use cases
- Transcript QA: Measure how close a machine transcript is to a human reference with WER.
- Translation review: Track edits between draft and final subtitle translations.
- Re-sync verification: Confirm a timing fix didn't change any wording.
- Framerate troubleshooting: Diagnose drift caused by a fps conversion.
The bottom line
Subtitle comparison is really two questions — did the words change, and did the timing change — and each has a clean answer. WER quantifies wording differences; timing analysis reveals offsets and drift. Together they turn a tedious side-by-side scroll into a couple of numbers and a line-by-line diff.
Compare your two subtitle files now
Upload both SRT or VTT files to DiffALL — get a WER score, timing drift analysis, and a line-by-line diff. Free, no install, no sign-up for your first comparison.
Compare subtitles now →