Handing a university's teaching footage to an outside company raises three reasonable anxieties: that the content will be altered in ways the academic did not intend, that the institution will lose control of its material, and that something will reach a student without anyone having approved it.
So here is the list. It is the same list we give every institution, and it is the one we expect to be held to.
What the pipeline changes
| We alter | What that means in practice |
|---|---|
| The picture | Exposure, white balance, contrast, stabilisation, framing and crop. The lecturer is not retouched, slimmed, aged or otherwise altered. |
| The sound | Room echo, hum, hiss and level. The voice is the original recording — processed, never regenerated. |
| The length and the cuts | Dead air at the start, long pauses, false starts and the ten seconds of finding the clicker. Segmentation into topic-length pieces. |
| What sits around the frame | Title card, speaker lower-third, institutional logo, background plate, opening and closing bumpers. |
| Text on screen | Captions and a transcript, generated from the spoken audio and corrected. |
| Added visuals | Where an idea benefits from it, an illustrative cutaway — stock, still or generated — shown to the faculty member in review. |
What the pipeline never touches
- The words. We do not rewrite, paraphrase, shorten by summarising, or reorder an argument. If a sentence is cut, it is silence or a false start, never a claim.
- The reasoning. The order in which a case is built is the teaching. It stays.
- The lecturer. No synthetic presenter, no voice cloning, no likeness reused in a video they did not record.
- The substance of an error. If a lecture contains a mistake, the produced version contains the same mistake, more legibly. We will flag it to you. We will not quietly correct a faculty member's academic content.
- Attribution. Names, designations and credits go on, not off.
Production makes a lecture easier to follow. It is not a review of the lecture, and we are not qualified to conduct one.
Ownership, footage and data
You own the output. Produced videos are the institution's material, without restriction. We do not claim a licence in your teaching content.
We do not use your footage to train anything. Your recordings are processed to produce your videos and for nothing else.
Source footage is deleted on request, and on a schedule by default. Tell us the retention period you need and we will work to it.
Students in the frame are a data question, not a production one. Where a recording captures identifiable students, the institution's own obligations apply — the DPDP Act in India, FERPA in the United States. We will blur, crop or decline the file, but we cannot give you consent you do not have.
Nothing publishes without a named approval. Every video returns for review. The log records the file, the effects applied, the reviewer and the date.
The review gate is the product, not the paperwork
The most common way an automated pipeline embarrasses an institution is a generated visual that attaches confidently to the wrong idea — an illustration of a legal concept from the wrong jurisdiction, a process diagram that inverts two steps. It looks authoritative, which is what makes it dangerous, and no amount of model quality removes the risk entirely.
The mitigation is a subject expert looking at it before a student does. We build that in and we do not offer a mode that skips it. It also happens to produce exactly the audit trail a quality assurance office needs at filing time, which is why we stopped treating it as friction.
What we will refuse
- Footage where we have been asked to remove a faculty member's credit.
- Any request to make a lecture appear to be delivered by someone who did not deliver it.
- Recordings of identifiable students where the institution cannot confirm it has the basis to process them.
- Changing what a lecture asserts, in either direction.
If any of the above is a problem for your project, it is better that we find out in week one than in week twelve.