How to QC Closed Captions & Subtitles: A 2026 Checklist
What does a caption QC pass involve?
A caption QC pass covers six areas: file and format validity, timecode and frame rate alignment, reading rate and line length against your style guide, positioning and safe area, transcription accuracy against the audio, and completeness across the full program. Automate the measurable checks; review the rest by eye.
Most caption rejections are not mysterious. They come from a short list of measurable violations that a style guide would have caught before delivery. The problem is that teams tend to QC by watching playback, which is good at catching glaring errors and poor at catching the ones that get files bounced.
This guide covers what to check, the numbers those checks run against, and where to draw the line between automated validation and human review.
The Caption & Subtitle QC Checklist
| Check | What to look for | Automatable |
|---|---|---|
| File integrity | Valid syntax for the target format, correct character encoding, no truncation | Yes |
| Timecode & frame rate | Start timecode matches the media, drop vs non-drop correct, no rate mismatch drift | Yes |
| Reading rate | Characters per second within the platform limit for the content type | Yes |
| Line length & count | Characters per row and rows per event within format limits | Yes |
| Event duration & gaps | Minimum and maximum on-screen duration, minimum frame gap between events | Yes |
| Illegal characters | Glyphs the target format cannot encode, smart quotes, stray control codes | Yes |
| Positioning | Text inside the safe area and clear of burned-in titles, lower thirds, and credits | Partly |
| Synchronisation | Events land with the audio, not ahead of or trailing the speaker | Partly |
| Accuracy | Text matches what is said, including names, terminology, and numbers | No |
| Completeness | Captions run the full program including credits, with no unexplained gaps | Partly |
| Non-speech information | Speaker IDs and sound effects present where a deaf viewer needs them | No |
The last four rows are where compliance actually lives. A file can pass every automated check and still fail the FCC's four quality standards — accuracy, synchronicity, completeness, and placement — because those are judgments about content, not syntax. See our FCC closed captioning requirements guide for how the Commission evaluates each.
What Actually Fails QC
In practice the rejections cluster. Reading rate violations are the single most common reason a subtitle file comes back, because they are invisible during playback — the text looks fine, it is simply on screen too briefly to read. Frame rate mismatches are the second: a 23.976 file conformed against 29.97 media drifts progressively, so the first act plays clean and the third act is a second out.
Then there is the type mismatch. Delivering translated subtitles where the spec asked for closed captions produces a file with no speaker IDs and no sound effects, which fails accessibility review even though every technical check passes. Our guide on closed captions vs subtitles vs SDH covers why the distinction matters at delivery.
Placement is the quiet one. Captions sitting over burned-in lower thirds or credits violate the FCC's placement standard and are easy to miss unless you review against the actual finished picture rather than an offline.
Reading Rate and Line Length Limits
These are the numbers your style guide should be enforcing. They differ by format and platform, and running the wrong set against a file produces confident false passes.
| Parameter | Broadcast captions (EIA-608) | Streaming subtitles / SDH |
|---|---|---|
| Characters per line | 32 | 42 (Latin script) |
| Lines per event | Up to 4 rows | 2 |
| Reading rate | Governed by style guide | 20 CPS adult, 17 CPS children (Netflix) |
| Minimum duration | Style guide dependent | 5/6 second (~833 ms) |
| Maximum duration | Style guide dependent | 7 seconds |
| Minimum gap between events | Style guide dependent | 2 frames at any frame rate |
Streaming figures reflect the Netflix Timed Text Style Guide. Other platforms vary — always validate against the spec you are delivering to.
Adopting a Style Guide
Do not write one from scratch. Netflix, Amazon, and the major broadcasters publish timed text style guides covering reading speed, line treatment, punctuation, and language-specific rules. The Netflix guide is the most thorough and is a reasonable baseline even for organisations that never deliver to Netflix.
What matters is that the guide is encoded as rules your software can check, not a PDF operators are expected to remember. A style guide that lives in a document produces inconsistent QC decisions; one that lives in the validator produces the same verdict every time.
Most teams end up maintaining several — one per delivery target — and selecting the right one at QC time. That is the correct pattern. A single universal style guide either fails valid broadcast files or passes invalid streaming ones.
How To Review a Caption File
Closed Caption Creator provides two tools that map onto the two halves of the checklist. Which one you reach for depends on where you are in the workflow.
Realtime Error Detection (RED)
When authoring (creating) or editing closed captioning, Realtime Error Detection (RED) alerts help notify users of issues as they appear. By configuring the Error Detection settings inside of Closed Caption Creator, it is easy to know when you've made a mistake, or something isn't quite right. RED alerts help identify and fix issues related to illegal characters, reading rate, duration, line lengths, and max characters.
Style Guide Validation
Style Guides are the review-stage tool: point one at a finished file and it scans every event against the configured rules, then lets you filter the failures by position, formatting, reading rate, frame gap, and duration. This is how you get a pass or fail verdict in minutes rather than a full playback pass.
Configure one Style Guide per delivery target and select the matching one at QC time. The QC and review feature page covers the full rule set.
The human pass
Automated validation cannot tell you whether a name is spelled correctly, whether a caption is sitting on top of a lower third, or whether the sound effect cues a deaf viewer needs are present. Budget for a review pass against the finished picture — not an offline — with the audio on.
Automating Caption QC
Once the measurable checks are encoded as a style guide, they can run without an operator. Our API accepts a caption file and a style guide expressed as JSON, and returns the violations — which means validation can sit at ingest, at delivery, or both, and reject bad files before a human ever opens them.
That is the shape most mature workflows converge on: automated gates catch the structural failures at volume, and human review is reserved for the accuracy and placement judgments that genuinely need eyes. Our three-part series on automating closed caption workflows walks through building this out, and part three covers API testing specifically.
Not a developer? Our engineering team can help integrate validation into an existing pipeline — see the work order API or custom integrations. For a wider look at QC tooling beyond our own, see our broadcast caption QC best practices article.
Frequently Asked Questions
What should a caption QC checklist cover?
Six areas: file integrity and format validity, timecode and frame rate alignment, reading rate and line length against your style guide, positioning and safe area, accuracy of the transcription against the audio, and completeness across the full program duration including credits.
What is the maximum reading rate for subtitles?
It depends on the platform. Netflix allows up to 20 characters per second for adult content and 17 for children's content. Reading rate violations are one of the most common causes of subtitle file rejection, so validate against your target platform's limit before delivery, not after.
How many characters per line are allowed?
EIA-608 broadcast captions are limited to 32 characters per row across a maximum of four rows. Subtitle and SDH formats generally allow 42 characters per line across two lines for Latin-script languages. Running the wrong limit against a file produces false passes.
Does passing technical QC mean a caption file is compliant?
No. A file can pass every format and encoding check and still fail FCC quality standards. The FCC grades accuracy, synchronicity, completeness, and placement, which are judgments about content rather than syntax. Automated validation catches structural problems; a human pass catches the rest.
Can caption QC be automated?
The measurable parts can. Reading rate, line length, character count, frame gaps, duration, illegal characters, and positioning are all rule-based and can be validated against a style guide via API at ingest or delivery. Accuracy against the audio and placement over on-screen text still need human review.
How many frames should separate two caption events?
Most streaming specs require a minimum of two frames between subtitle events at any frame rate. Netflix also sets a minimum event duration of 5/6 of a second, roughly 833 milliseconds, and a maximum of seven seconds. Check your target platform's spec, as these vary.
Want to run this checklist against your own files? Start a free trial or talk to our team about automating validation in your pipeline.