Subtitle & Caption Checker
Free · No sign-up · Runs in your browser — your file never uploads.Every problem cue, with its timestamp, ready to fix.
The rules it checks
- Reading speed — characters per second per cue; over ~20 CPS flagged (over ~25 = severe).
- Line length — lines over 42 characters; more than 2 lines per cue.
- Cue duration — under 0.8s (flashes) and over 7s (hangs).
- Timing hygiene — overlapping cues and gaps under 80ms, which players render as flicker.
- Format validity — unparseable blocks reported with their position, not swallowed.
- Profanity flags — cues matching a built-in list of strong profanity terms, flagged with timestamps for review. List-based flagging only: context, quotes and slang can false-positive, so a flag is a prompt to review, never a declared violation.
How the checker reads your file
The file is read as plain text in this tab and split into blocks at blank lines. In each block the checker finds the timing line — the one containing --> — and parses its two timestamps: SRT writes them with a comma (00:01:02,500), VTT with a dot, and the hours are optional. The lines after it are the cue's text. A VTT header and NOTE blocks are skipped; a block with no readable timing line is reported as an error, never silently dropped.
Why each rule matters
- Reading speed — a cue's characters divided by its seconds on screen. Speech runs faster than reading — above all for deaf and hard-of-hearing viewers, who get everything from the text. Past roughly 20 characters per second people start dropping words; past 25, many rewind or switch captions off. YouTube's auto-translate also works cue by cue, so cues that break at natural pauses translate more cleanly.
- Line length — players wrap a long line wherever they choose, often mid-phrase; on a phone it shrinks the type or covers the picture. Two lines of at most ~42 characters stay inside the safe area on any screen. A third line starts hiding faces and action.
- Cue duration — under ~0.8 seconds a cue is a flash nobody can read; over ~7 seconds the text sits unchanged across cuts and new sentences, so viewers re-read it or assume captions have stalled. Long cues are usually two sentences that want splitting at the pause.
- Overlaps and gaps — when a cue starts before the previous one ends, players disagree: some drop a cue, some stack both, and either way the fix is re-timing by hand later. A gap under ~80ms is too short to register as a break, so the text flickers instead — make it zero, or a real gap.
- Profanity flags — whole-word matches against a built-in list, reported with timestamps. YouTube's advertiser-friendly guidelines treat strong language, above all in the opening seconds, as a limited-ads risk. The list can't hear context — a quotation or a lyric may be deliberate — so a flag is a review prompt, never a verdict.
What this tool cannot tell you
- It checks timing and shape, not translation quality or accuracy — a perfectly-timed wrong subtitle passes.
- The thresholds are broadcast conventions, not YouTube rules. Fast CPS in a fast-cut Short is a different judgment than in a lecture.
- It never hears the audio: it cannot tell whether a cue lines up with the words actually being spoken — only whether the cues are consistent with each other.
Worked example: one bad cue, fixed
An invented example — labelled as such — scored by exactly the rules above:
1 00:00:04,000 --> 00:00:04,600 And this is where the whole experiment completely fell apart on us
That cue packs 66 characters into 0.6 seconds: about 110 characters per second. It also breaks the 42-character line limit and the 0.8-second minimum duration — three flags from one cue. Split at the natural pause and give each half real time:
1 00:00:04,000 --> 00:00:06,000 And this is where the whole experiment 2 00:00:06,000 --> 00:00:07,900 completely fell apart on us
The first cue now runs at 19 characters per second and the second at about 14 — inside every limit, and each line fits a phone screen.
Frequently asked questions
What makes subtitles 'bad' for viewers?
Speed and shape. Text that flashes by faster than people read (over ~20 characters per second is a strain for most audiences), lines longer than ~42 characters that wrap unpredictably, cues shorter than a second, and cues that overlap or nearly touch — each one quietly pushes viewers to turn captions off or leave.
SRT or VTT — does it matter?
YouTube accepts both (and others). VTT supports positioning/styling; SRT is the universal plain format. This checker reads both, using the same timing and readability rules.
Why 42 characters per line?
It's the broadcast subtitling convention (BBC/Netflix style guides cluster around 40–42) because it keeps two lines readable on any screen without covering the picture. It's a guideline, not a YouTube rule — we flag, you decide.
Do captions affect my video's reach?
Captions make videos watchable muted, in noisy places, and in other languages via auto-translate — all real retention and audience effects. YouTube also indexes caption text. What captions don't do is magically rank a video; they're an accessibility and completion feature first.
My auto-captions are full of errors. Fix or replace?
Replace for anything you care about. Auto-captions mangle names and technical terms, and YouTube shows the errors to exactly the viewers who need captions most. Edit the auto-captions in Studio or upload a corrected file — this checker will audit the timing either way.
Do you upload my subtitle file?
No. It's parsed in your browser and forgotten when you close the tab.
Related tools
- Video Analyzer — the video itself: hook, silence, pacing.
- Metadata Viewer — chapters and tags live in the description; check them there.
CreatorChecks is not affiliated with YouTube or Google.
Was this result accurate?
Thanks — feedback like this helps us catch errors.
Could not send — please try again.