Do I need to caption every minute, or can I skip intros and outros?

Caption everything that has speech: spoken intros, the Q&A at the end, questions from the audience. Silent title cards and music-only transitions don't need captions, though a [music] tag tells a viewer they aren't missing anything. In practice the AI transcribes the whole recording anyway, so captioning everything is less work than deciding what to skip.

AI Subtitles for Online Courses: Accessibility Rules and a Workflow for Lecture Libraries

Why courses get captioned

Captions let deaf and hard-of-hearing students follow a lecture, but they help far more people than that. Students working in a second language can read along, people watching on a train or next to a sleeping child can follow with the sound off, and a transcript makes it possible to search a two-hour recording for the moment a topic came up.

Which rules apply

This is a summary, not legal advice. Your institution's accessibility office will know which of these applies to you.

United States

ADA Title II covers public institutions, including public universities. Title III covers private organizations open to the public. Courts have disagreed on how far Title III reaches for services that exist only online.
Section 504 of the Rehabilitation Act applies to programs that receive federal funding, which includes most US colleges.

European Union

The European Accessibility Act has applied since June 2025 to a defined list of products and services. Whether your courses fall under it depends on how they are sold and delivered, so check rather than assume.

The technical standard

WCAG 2.1 Level AA is the standard most policies point to. Success Criterion 1.2.2 requires captions for prerecorded audio in synchronized media.
WCAG doesn't set an accuracy percentage, but captions have to carry what was said. Unreviewed auto-captions often get names, terms and numbers wrong, which is why the review pass below matters.

What it costs with Picute

Picute charges for transcription only: 1 credit is 1 minute, rounded up per job. Burning subtitles into an MP4 doesn't use credits on any plan.

A semester of 40 one-hour lectures is 2,400 minutes, so 2,400 credits. The Pro plan ($29/month) includes 900 credits a month and Premium ($49/month) includes 1,500. Every new account gets 10 credits once, enough to test a 10-minute excerpt before you commit. On the free plan, video exports carry a Picute watermark.

Workflow for captioning a course library

Step 1: Get the recordings in

Upload the lecture files, pick them from Google Drive, or, if the lectures are on YouTube, paste up to 10 links at once. There is no length cap, so a three-hour seminar goes in as one file.

Step 2: Set the language and prepare a term list

Choose the spoken language yourself instead of relying on detection, especially for non-English lectures. Picute transcribes 60 languages.
Pull a list of names and terms from the syllabus: lecturer and guest names, institutions, reagents, case names, acronyms. You'll check the transcript against it in the next step.
Turn on speaker labels if the recording includes discussion or audience questions.

Step 3: Review each lecture

Correct proper nouns against your term list.
Fix domain terminology (formulas, citations, jargon).
Check every number and date.
Leave filler words alone unless you publish the transcript as text.

Step 4: Export and upload

Export SRT or VTT for each lecture. Most LMS video players accept one or both; check your LMS's help page for caption uploads before processing the whole library.
If your player needs the other format, the free SRT to VTT converter handles it in the browser.

Step 5: Check a few lectures in the real player

Open three to five lectures in the LMS itself and look for:

Captions that appear when the words are spoken
Readable text in the player's default styling
Special characters (formulas, accented or non-Latin names) rendering correctly

Language considerations

English lectures. Domain vocabulary is where most corrections happen.

Korean, Japanese, Chinese, Thai and Vietnamese lectures. Results vary between tools, so transcribe one sample lecture and review it before running the whole library.

Lectures that mix languages. Short switches into another language are common in lectures. Check those passages closely during review.

Special content types

Math and formulas

Spoken formulas come out as words ("x squared plus y squared equals r squared"). For a published transcript, replace them with notation during review.

Lab demonstrations

Add tags such as [screen shows X] or [demonstrates technique] where the meaning depends on what's on screen. Speech recognition only writes down speech, so these have to be added by hand.

Q&A with the audience

Speaker labels help separate the lecturer from students. Questions asked away from the microphone may be unintelligible; mark them [inaudible] rather than guessing.

Guest lecturers

Speaker labels are generic, so rename them to the guest's name during review.