How to Turn a Sonix Transcript into Subtitle Files That Pass QC
Sonix exports SRT and VTT files, but the raw export usually fails platform checks on three counts: lines longer than 42 characters, reading speed above 20 characters per second, and captions that straddle a cut. All three are fixable inside the Sonix editor before export, which is faster than fixing them in a subtitle tool afterwards.
This is the checklist I run on every video before uploading captions.
Why does a raw Sonix SRT get rejected?
Most platforms run automated caption QC on line length, reading speed and segment count. Sonix splits by sentence, which produces segments that read well but violate the line-length rule whenever a sentence is long.
The export also carries speaker labels inline if you left them on, and those go straight into the caption text. That alone gets a file flagged as unreadable in some pipelines.
Check 1: line length under 42 characters
Two lines maximum, 42 characters each is the usual standard for broadcast and most streaming platforms. YouTube tolerates slightly more but readability drops.
In the Sonix editor, split long segments at a natural clause boundary. Do not split mid-phrase — a caption that breaks “the revenue was up / 14 percent” forces the eye to jump. Break before the preposition instead.
At an average of 150 words per minute, expect roughly 12 to 18 captions per minute of speech. If your file has 8, the segments are too long.
Check 2: reading speed under 20 characters per second
Reading speed is where automated QC fails most often. Fast speakers produce segments that change too quickly for the eye.
Sonix shows duration per segment. Multiply characters by 1 divided by duration in seconds to get CPS. Anything above 20 needs one of three fixes: merge with the next caption to extend duration, drop filler words (“you know”, “like”, “I mean”), or shorten the phrasing.
Dropping fillers is legitimate for captions even when the transcript keeps them. A transcript is a record; a caption is a reading aid.
Check 3: speaker labels out, punctuation in
Turn speaker labels off in the caption export. If the video shows who is speaking, labels are redundant; if it does not, use upper-case name prefixes only for the first appearance of each speaker.
Punctuation should stay. Platforms render it, and it measurably improves comprehension for deaf and hard-of-hearing viewers. Sonix auto-punctuates; review sentence ends, since run-on segments are where it slips.
How do you handle the segments that do not fit either rule?
Some passages simply cannot satisfy both. A fast technical explanation at 22 characters per second has no filler to cut.
Options, in order of preference: extend the caption 0.3 seconds past the speech (viewers tolerate slight lag), simplify the wording using the same meaning, or split into three short captions with one second of overlap. Overlapping captions are legal in SRT and readable in most players.
Never shorten by deleting meaning. A caption that says something different from the audio is worse than a fast one.
What else is worth checking before upload
- Numbers and names. Auto-transcription gets proper nouns wrong most often; a custom vocabulary list before transcription cuts this sharply.
- Timing at cuts. Captions spanning a hard cut look broken. Snap them to the cut.- Non-speech audio. Add
[laughs]or[music]where relevant; automated QC does not require them but reviewers do.## When is the SRT approach the wrong choice?If you are captioning a long series, a dedicated captioning service or a platform’s own auto-captions with manual review will be cheaper per hour than editing Sonix output segment by segment.If you need burned-in captions styled to a brand, export plain text and build the styling in your video editor — styling inside SRT is not supported by most platforms.For a one-off interview with a client deadline, the Sonix export plus this three-check pass takes about 15 minutes for a 10-minute video, which is hard to beat.## Quick Summary- Raw Sonix SRT exports fail QC on line length, reading speed and inline speaker labels.- Keep captions to 42 characters per line, two lines maximum.- Target under 20 characters per second; merge, trim fillers or rephrase to get there.- Turn speaker labels off and keep punctuation on for comprehension.- Fix in the Sonix editor before export — it is faster than repairing the SRT later.
