Free Creator Toolkit
Publishing

Fix SRT and VTT files: formatting, timing and readable captions

Diagnose rejected subtitle files, convert SRT and VTT safely, correct a constant timing offset, and check the exported captions against your video.

4 min readBy Free Creator Toolkit Editorial TeamUpdated

Start with the symptom and keep the original

A rejected subtitle upload, captions appearing late, and captions disappearing too quickly are different problems. If the editor cannot read the file, check its format and structure. If it imports successfully but the words arrive at the wrong moment, check synchronization. If timing matches the speech but the text is hard to read, review cue length and duration.

Keep an untouched copy before making changes. Use the tools below with a subtitle file you already have; they do not listen to a video or transcribe missing speech. Selecting a local caption file processes its text in your browser. A repaired file still needs review, especially when the report identifies text or cues that could not be recovered.

Check the text format before repairing timestamps

An SRT file is plain text, not a word-processing document renamed with an .srt extension. YouTube's supported-format guidance specifies plain UTF-8 for basic SubRip captions and says SRT styling is not recognized. If accented characters are corrupted, return to a correctly encoded source; changing a filename cannot restore letters that were already replaced during an earlier export.

A synthetic SRT example has four parts: cue number 1 on its own line; 00:00:01,000 --> 00:00:03,500 on the next line; the text Welcome to this example. below it; and a blank line before cue 2. Use the SRT validator to identify numbering, separator, and timestamp problems. An end time earlier than its start requires a decision about the intended timing, not an arbitrary replacement.

Convert the format without assuming all styling survives

WebVTT begins with WEBVTT and uses a period for fractional seconds: the same example becomes 00:00:01.000 --> 00:00:03.500. Its format also supports cue settings, comments, style blocks, and regions. Choose SRT or VTT according to the destination's requirements. Renaming .vtt to .srt leaves the contents unchanged and is not a conversion.

Review the converter's diagnostics and exported file when the source contains positioned captions, speaker markup, or styling. SRT and VTT do not represent every feature in the same way, and a basic conversion may lose presentation information. Keep the original VTT if its layout matters. Plain-text extraction deliberately removes timing; use that output for transcript editing, not as a replacement timed-caption track.

Separate a constant delay from gradual timing drift

Check one spoken phrase near the beginning and another near the end. If every caption appears 1.5 seconds too early, enter 1.5 in the time shifter's seconds field to move it later. If it appears 1.5 seconds too late, enter -1.5. For a synthetic cue from 00:00:04,000 to 00:00:06,000, the positive 1.5-second offset produces 00:00:05,500 to 00:00:07,500.

An offset cannot fix captions that become progressively further out of sync. That can happen when a track belongs to another edit or timing scale. Return to the editor and align multiple points. When moving cues backward, inspect the first cues for clipping at zero and any changed duration. Review overlaps individually: simultaneous speakers can be intentional, whereas an accidental overlap can hide dialogue.

Use reading speed as a review signal

The reading-speed checker highlights cues that deserve attention; it does not certify accessibility or replace listening. As a simple planning example, 60 visible characters shown for three seconds require 20 characters per second. Whether that is comfortable depends on the audience, language, vocabulary, and the visual work happening at the same time.

When a cue is dense, consider a shorter accurate phrasing, a break at a natural phrase boundary, or more display time where the audio allows it. Do not extend a caption across unrelated speech merely to improve a number. Retain meaningful speaker identification and sound descriptions when preparing captions, and review names, specialist terms, and punctuation against the recording.

Inspect the downloaded result in its destination

Download a separate result and compare the reported cue count with the original, accounting for every omission. Read the diagnostics before accepting repairs. Then load the result alongside the final video and inspect the opening, closing, transitions, multilingual characters, and any flagged cue. A structurally valid subtitle file can still contain incorrect words or belong to the wrong video version.

Keep the reviewed timed-caption file and any plain-text transcript as separate deliverables. Name them with the video version and language so a future edit does not silently reuse outdated timings. The practical order is validate, convert if needed, synchronize, review readability, and finally check the actual player. Repeat only the step that addresses the remaining problem.

Primary references

Sources and verification