Use case · Higher education

University lecture capture captions: ADA Title II scope, Kaltura/Panopto flow, and the back-catalog problem

Public universities are squarely in scope for the ADA Title II caption deadline that went live 2026-04-24. Years of lecture-capture archive — Kaltura tenants, Panopto installations, Echo360 libraries — are in scope retroactively where the content is still being served to enrolled students. A typical R1 university lecture-capture archive is in the tens of thousands of hours; a mid-sized state university is in the low thousands. Auto-captioned at ~85% accuracy, none of it passes a sample audit. Here is the realistic retrofit playbook, why glossary-awareness is the crux for academic content, and the upload flow into Kaltura and Panopto.

TL;DR

Public-university lecture capture is in ADA Title II scope as of 2026-04-24. Kaltura and Panopto both accept SRT/VTT/TTML caption tracks via UI and API. The back-catalog is too large to caption manually but too small (per-program) to merit a Verbit/3Play enterprise contract. The right path is glossary-aware captioning by department: each program's term list (chemistry, biology, CS, law, history) feeds the caption model so degree-program proper nouns land right on first export. Prioritise by enrolled-student usage; deprecate or archive content that no current cohort streams.

What ADA Title II actually requires for lecture capture

ADA Title II covers state and local government entities, including state-funded public universities and (in many states) the city or county college systems. The 2026-04-24 deadline reset the long-deferred expectation that captioned video on official educational programmes is mandatory, not aspirational. The relevant compliance frame is WCAG 2.1 AA, which inherits the SC 1.2.2 requirement for synchronized captions on prerecorded content (see our SC 1.2.2 page for the exact language).

The audit posture matters. A complaint to the DOJ Civil Rights Division about an inaccessible lecture-capture archive triggers a process that asks for the institution's caption policy, recent audit results, and a remediation plan. "We have YouTube auto-captions" is not a passing remediation plan; YouTube auto-captions cluster in the 80–90% accuracy band on subject-dense academic content, well below the 99% bar.

The back-catalog scale problem

A typical R1 university produces 5,000–15,000 hours of lecture capture per academic year across departments. Multi-year archives in Kaltura and Panopto run into the tens of thousands of hours. The cost math at human-captioning rates ($1.50–$4.00 per minute for terminology-aware captioning) is large six to seven figures. The cost math at general-purpose AI captioning is fractional per-minute but produces the 85%-accuracy garbage that fails an audit.

The realistic prioritisation:

Currently-streamed content first. Pull the streaming-event log from Kaltura or Panopto for the last academic year. Modules with non-zero current-cohort views are in scope; modules with zero views can be archived or deprecated, removing them from compliance surface.
Currently-enrolled programmes second. Caption the active back-catalog for currently-running programmes. Discontinued programmes have a different governance posture; consult your institutional accessibility office.
New captures captioned on ingest. All future lecture captures should run through a captioning-on-ingest pipeline (GlossCap or otherwise). The back-catalog problem doesn't recur.

Why glossary-awareness matters more for academic content

Academic content is dense with discipline-specific proper nouns and terms-of-art that no general speech model has been pretrained on:

Chemistry. IUPAC names of reagents, common-name aliases, lab-equipment proper nouns. Tetrahydrofuran, dichloromethane, N,N-dimethylformamide.
Biology and biomedical. Drug names, anatomy, pathway names. CRISPR-Cas9, phosphatidylinositol-3-kinase, cholecystokinin.
Computer science. Tool names, library names, algorithm names. Kubernetes, PyTorch, RoBERTa, Levenshtein.
Law. Case names, doctrines, foreign legal terms. Marbury v. Madison, stare decisis, res ipsa loquitur.
History. Place names with non-anglicised pronunciations, period-specific titles, primary-source proper nouns.

Each department's term list is small (a few hundred to a few thousand terms) but high-leverage: these are exactly the words a student needs to spell correctly to study, search, or reference downstream. A glossary-biased decoder gets these right on first pass; a general decoder writes phonetic guesses that fail every audit and every comprehension check.

The Kaltura flow

Kaltura is the dominant LMS-adjacent lecture-capture platform in higher ed. Caption attachment flow:

List entries by department or course. Use Kaltura's media.list API filtered by category or owner; the entry-list export goes to a working spreadsheet.
Caption with department-specific glossary in GlossCap. One workspace per department; one glossary per workspace. The chemistry department's IUPAC list, the CS department's tool list, the law school's case-name list — each scoped to its own captioning batches.
Bulk-attach via REST. The Kaltura Caption API takes a two-call pattern (caption_captionasset.add + setContent); see our Kaltura captions page for the exact flow.
Verify on a sample course. Pull 10 sampled entries from a current-term course, open in MediaSpace as a non-admin, confirm CC button shows the right track and timing looks right.

The Panopto flow

Panopto's caption upload is on the session edit page (Settings → Captions). The platform supports SRT and VTT upload directly, and provides a Captions API for bulk attachment programmatically. The retrofit pattern mirrors Kaltura's: enumerate sessions per folder/course, caption with a department glossary in GlossCap, attach via Panopto's API, verify on a sample. Panopto also runs an in-platform ASR service ("Auto-captioning") which produces the same 85%-accuracy pattern that fails an audit on subject-dense material; the right move is to disable in-platform auto-caption on retrofit sessions and replace with the GlossCap export.

Compliance practical posture

Most institutional accessibility offices will accept a documented retrofit plan with a defensible timeline if a complaint is filed before retrofit completes. The plan needs:

Inventory of in-scope content.
Prioritisation rationale (currently-streamed first, etc.).
Captioning workflow with documented quality bar (glossary-aware, departmental review, sample-audit cadence).
Deadline for full retrofit.
New-capture pipeline that ensures the problem doesn't recur.

The GlossCap workflow checks every box in this list: the glossary is auditable, the per-batch output is sampleable, the per-customer model improves with each correction, and the captioning-on-ingest mode handles new captures.

See pricing