Forty-two chairs. That is the education room at a mid-sized regional museum I visited in March, and it is why their Tuesday evening curator talks close registration about ninety minutes after the announcement goes out.

The talk that night was about a small painting that had recently been reattributed. The curator had spent two years on it, and she spoke for thirty-five minutes without notes before taking questions for another fifteen. At nine o'clock the chairs went back on the trolley. There was no recording, nothing on the website, and when the exhibition came down in June the only surviving trace of two years of research was a wall label of about ninety words.

We do this constantly. Most of us have stopped noticing.

What goes missing when the talk ends
Museum education is built around live delivery, and there is nothing wrong with that instinct. A person standing in front of the object is what we are for.

The problem sits on either side of the hour. Interpretation gets written for one exhibition, delivered five or six times, then dissolves when the show travels. A docent who has given the same twenty-minute talk since 2019 retires, and the department finds out nobody ever wrote it down. A school group turns up cold because the teacher had nothing to prepare them with.

The waitlist is the version of this you can see. The version you cannot see is the interpretive work a department produces every year and keeps almost none of.

Recording the talk is not the same as making a lecture
Put a camera at the back of the room. Most of us have tried it, and the results are usually unusable for reasons that have little to do with equipment.

Gallery acoustics are bad. Curators walk. The camera captures the back of visitors' heads and a painting reduced to eighty pixels of relevant surface, and whatever the curator was pointing at, the viewer cannot see. Pointing is roughly half of what happens in a gallery talk.

Then there is rights, which no amount of equipment fixes. A license for a loaned work usually covers wall display and the print catalogue, and says nothing useful about a video that will sit on your website for three years. Filming in a gallery hung with loans creates a distribution question that nobody in the room has the authority to answer.
So the recording either does not happen or happens badly. The department goes back to live delivery, because live delivery is the thing that works.

Working from what we already write
Here is the part I think we undervalue: the material for a proper lecture already exists in most departments.

The checklist PDF. The catalogue essay. The docent training deck from the last rotation. The curator's talking points, which are almost always a Word file with the dates in bold. The interpretive plan that took four meetings to settle. We generate a great deal of written interpretation and then use each piece exactly once.

What has been missing is a workable route from that pile of documents to something a person can watch. An AI lecture video maker like Leadde is built for that specific gap. You hand it the deck, the PDF, the Word file, or pasted text, and it drafts an outline, lays out the on-screen scenes, and generates the narration.

Before anyone dismisses this as automated filler, two of the settings matter. You can specify the audience and the speaker's background, and the script changes accordingly. Notes written for a curatorial colleague produce something quite different from the same notes aimed at a family audience or a Year 10 class. You also choose the narrative style, and for our purposes Analytical and Explanatory are the useful ones rather than the more persuasive registers. Length works the same way, from a Summary version of one to three minutes on a single object up to In-depth at five to seven for a full thematic argument.

If your department has a real archive, look at the Knowledge Base first. You can batch upload source material ten files at a time and build a searchable body of your own interpretation to work from. Fifteen years of catalogue essays is a very different starting point from an empty page.

Where this fits in a department
Docent and volunteer training is the strongest case, and I would start there rather than anywhere public-facing. Volunteer turnover is high, training cohorts are annual, and the content barely changes between years. A department that builds twelve short lectures on its permanent collection has fixed a recurring problem instead of a one-off. New volunteers watch before their first shift rather than shadowing someone who learned it from someone who learned it from someone.

School group preparation comes second. Teachers arrive with students who have never been asked to look at a painting for longer than four seconds, and a five-minute lecture on how to read a history painting changes what happens when the group gets off the bus. Attach it to the booking confirmation and it costs the teacher nothing.
Multilingual access is where I would spend the effort if your visitor numbers justify it. A finished video can be translated into another language, script and on-screen text together, across 88 languages and 175 dialects. A translated wall label gives a Korean or Portuguese speaker a compressed summary. This gives them the argument.

Traveling exhibitions deserve a thought as well. A show that goes to three venues usually gets its interpretation rebuilt three times, and there is no good reason for that.

Most departments I know contain nobody who wants to appear on camera. The platform has over 200 built-in presenters, and you can generate one from a single photograph, so a curator who will not stand in front of a lens can still hold a series together.

Shared videos also carry a chat panel, so a viewer can ask a question about what they just watched and get an answer back. I would not oversell this. It is not a curator, and I would not want it described to a school as one. But the questions people actually ask after a talk are repetitive, and half of them are about dates and materials, which is exactly the sort of thing that does not require a curator's evening.

One more thing changes how you evaluate the work itself. Shared videos carry an analytics view with completion rate and average watch time. Attendance tells you forty-two people came. Completion rate tells you two-thirds of viewers stopped at minute four, which is a fact about your interpretation that no live talk has ever handed you.

Where it falls short
Several of the limits here are particular to our field, and they are worth stating plainly.

Rights come first. The tool does not resolve your reproduction permissions. If the license for a loaned work does not cover video distribution, generating the video creates an obligation rather than removing one. Sort that before you make anything.

Close looking does not transfer. Standing at the correct distance from a canvas and watching the surface change as you move is the thing our institutions exist to provide. A video is preparation for it or a memory of it. Do not let anyone in your organization describe it as a substitute, because that argument has a way of resurfacing during budget season.

Give the generated presenters a few seconds of real attention and something in the delivery gives way. Anything carrying institutional voice, a director's statement, an acquisition announcement, a message about a repatriation claim, should come from a real person. Who is speaking is part of what makes those messages credible.

The output inherits whatever is wrong with the input. A source document that is loose about attribution or hedged about dating will produce a script that is loose in the same places and confident about it. Someone with subject knowledge has to read the generated script before it goes anywhere. Treat it as a first draft from a capable research assistant who has never seen the object.

The platform expects a desktop and a Chromium browser, which is worth checking against whatever machines your institution actually hands out.

A reasonable pilot
Take one talk you have given more than three times, because that means the argument is already settled.

Run the existing notes through the free tier at Balanced length, which lands around three to five minutes. Ask the curator who wrote the notes to read the generated script and mark what is wrong. Fix it. Put it on the exhibition page and send it to your docent cohort.

Look at the completion rate in October. If people watched it through, and if your docents came into training with better questions than last year's group, you have a repeatable way to hold on to interpretation that currently disappears every time a show comes down.

The forty-two chairs are not going anywhere. What the curator said in front of them does not have to end with the evening.