ADA Title II and your university's video: what the rule requires, and the date that moved
This is a plain reading of a published rule, not legal advice. If your institution has counsel, they get the last word.
If you run communications at a public university, you have probably been sent a vendor’s compliance page about ADA Title II, and it probably wanted a meeting. Here is the same information without the meeting.
Who this applies to
The Department of Justice’s web accessibility rule under Title II of the ADA covers state and local government entities — which includes public universities and colleges, along with cities, counties, school districts and special districts. Private institutions are not covered by this particular rule; they have separate obligations under Title III that this rule does not change.
Two things about the rule matter more than anything else in it: the standard, and the date.
The standard is WCAG 2.1, Level AA. Not “accessible in spirit” — a specific, published, testable version of the Web Content Accessibility Guidelines, applied to your web content and mobile apps.
The date moved. The original deadlines were April 2026 and April 2027. On 20 April 2026 the DOJ issued an interim final rule extending both by one year:
| Entity | Compliance date |
|---|---|
| Population of 50,000 or more | 26 April 2027 |
| Population under 50,000, and any special district government | 26 April 2028 |
A large state university system falls in the first row. If you have been working to an April 2026 date, you now have a year you did not have. If you have been treating “it got delayed” as “it went away”, you have about eight months.
What WCAG 2.1 AA actually requires of video
This is the part the vendor pages usually skip, because it is short and it is checkable. Three success criteria decide whether your video passes:
- 1.2.2 Captions (Prerecorded) — Level A. “Captions are provided for all prerecorded audio content in synchronized media, except when the media is a media alternative for text and is clearly labeled as such.” Every recorded lecture, every promo video, every panel discussion on your site.
- 1.2.4 Captions (Live) — Level AA. Live synchronized media needs captions too. Commencement streams, live-streamed lectures, a broadcast board meeting.
- 1.2.5 Audio Description (Prerecorded) — Level AA. Prerecorded video needs audio description of what is shown but not spoken. This one is routinely forgotten and it is at the level you have to meet.
Note what is not in there: nothing says the captions have to be produced by a human, and nothing names a vendor. What the criteria require is that captions exist, that they are synchronized, and that they convey the audio. A caption track that says the wrong words, or shows the right words two seconds late, does not convey the audio — which is the honest reason auto-captions straight off an upload usually fail, and it is a quality problem, not a licensing one.
What you do not have to caption
The rule has real exceptions, and reading them carefully is the cheapest thing you will do this year. From the DOJ’s own summary:
- Archived web content — but all four conditions have to hold: created before the compliance date, kept only for reference or recordkeeping, stored in a designated archive area, and unchanged since it was archived. A lecture recording that students are still being pointed at is not archived, whatever folder it is in.
- Preexisting conventional documents — Word, PDF, presentation and spreadsheet files that were on the site before the compliance date. The exception drops away for any document currently used to access a government service.
- Third-party content — content posted by members of the public or by unaffiliated parties. Content posted by the institution itself is not third-party content, wherever it is hosted.
- Individualized password-protected documents — files about a specific person, property or account.
- Preexisting social media posts — posts made before the compliance date.
Read together, these say something practical: the obligation is heaviest on what you publish from here on, and on the older material you are still actively directing people to. An inventory that separates “actively linked” from “genuinely archived” will shrink the job more than any tool will.
What compliance costs per hour
This is where the numbers get strange, because the same deliverable is sold at prices that differ by two orders of magnitude.
| Route | Price per minute | Per hour of video |
|---|---|---|
| 3Play Media, 10-business-day turnaround | $1.90 | $114 |
| 3Play Media, 2-business-day | $2.60 | $156 |
| 3Play Media, 2-hour turnaround | $6.00 | $360 |
| Rev, human captions (English) | $1.99 | $119 |
| Rev, AI captions | $0.25 | $15 |
| Speecho, pay-per-use | $0.014–$0.021 | $0.83–$1.25 |
Those are not the same product, and pretending otherwise would be the sales pitch this post is trying not to be. The $114-an-hour tier includes a human who listens to the whole file and is accountable for the result. The $1-an-hour tier is a machine transcript that you have to read.
The question is which one your actual workload needs. If your office publishes four hours of video a month — a couple of lecture recordings, a press conference, a donor video — the human-service route is around $5,500 a year and the machine route is around $50 plus your own time. If you publish four hundred hours a year and none of it is checked by anyone, no tier will save you, because the failure will be quality, not budget.
The workflow that actually passes
For a communications office with more video than budget, this is the shape that works, and the third step is the one that is not optional:
- Generate the transcript. Upload the recording, get back a timestamped transcript and an
.srtor.vttfile. On Speecho an hour costs about a dollar and takes minutes, with no subscription, which matters when your video volume is lumpy — three hours in commencement week and nothing in July. - Check the timing. Machine captions are usually accurate and still uncomfortable, because raw segments ignore reading speed. Our free subtitle readability checker measures your file against the broadcast conventions — 17 characters per second, 42 characters a line, a minimum duration per cue — and tells you which cues fail. The rules behind those numbers are worth ten minutes of your time once.
- Have a person read it. Names, acronyms, department titles, the visiting speaker’s institution: these are exactly what speech recognition gets wrong and exactly what your audience notices. Ten minutes of proofreading per hour of video is the difference between a caption track that conveys the audio and one that technically exists. There is no tool, ours included, that removes this step from a compliance obligation.
Then add the audio description pass for anything where the picture carries information the audio does not — a demonstration, a chart on screen, a silent b-roll sequence.
A one-page checklist for the office
- Confirm which deadline you fall under: 26 April 2027 if your entity serves 50,000 or more, 26 April 2028 otherwise.
- Inventory published video, split into actively linked and genuinely archived — and be honest about which folder each one is really in.
- Caption everything actively linked, prerecorded first (1.2.2 is Level A), then live streams (1.2.4).
- Add audio description where the visuals carry unspoken information (1.2.5).
- Write down who proofreads, and how long that takes. A workflow without a named human is a workflow that fails on accuracy.
- Keep the caption files.
.srtand.vttare plain text, they are yours, and they outlive whichever platform you are on this year — our free converter moves between the formats without uploading anything.
The rule is not asking for a captioning contract. It is asking for captions that are accurate, synchronized and present. How you get there is up to your office, and for most offices the honest answer is a cheap transcript plus a careful reader — not a purchase order.
