Narrated Video, or a Lesson That Answers Questions? A Teacher's Guide to Reusing Your Slides#
Starting from slides that already exist#
Most advice about AI for teachers starts from a blank page: generate a lesson plan, generate a quiz, generate slides. That's useful the first time you teach something. It's beside the point the fourth time.
By then the slides exist. It's been through a term, maybe three. The slides are fine. What isn't fine is that you stand up and deliver roughly the same forty minutes again, and afterward the same six students email you the same three questions.
There's a whole category of tools aimed at exactly this — take the slides you already have, add narration, hand something back you can send to students. They mostly split into two routes, and the routes lead to different places. Almost nobody says so, because each tool sells the route it happens to be on.
So here is what each one actually hands you at the end.
Narrated video: you get a file#
What you get#
You upload the slides. The tool writes a script slide by slide, reads it in a synthetic voice, syncs it to the slides, and exports a video file. Some tools use your speaker notes if you wrote any. Some embed the audio back into the original file instead of exporting separately.
The output is an MP4. You own it. You can put it in your LMS, drop it in a shared drive, hand it to a colleague, keep it after you stop paying for the tool.
Where it works well#
It's the right answer when the material is settled and the delivery matters more than the conversation — a lab safety briefing, a module introduction, an orientation you give every September. Students who watch on a commute, or somewhere with bad Wi-Fi, need a file rather than a page. So does a colleague reusing your material in a course you have nothing to do with. And when someone above you wants something for an accreditation folder, a link into someone else's platform is not a file.
Video is also finished, in a way a hosted page never is. It doesn't stop working when a service shuts down.
Where it runs out#
A video answers the questions you anticipated while writing the script. Student questions don't arrive on that schedule. They arrive at 11pm the night before an exam, on slide 12, about the step you thought was obvious.
So the video gets watched, and then the email arrives anyway. You've automated the part that was already working — the explanation — and left the part that actually consumes your evenings exactly where it was.
The second problem is edits. Change one slide and you re-render, re-upload, and chase down every place the old copy is still sitting. Anyone who has taught the same course for three years knows there's a version circulating that's two terms out of date.
A presentation that answers: you get a link#
What you get#
Same starting point: you upload the same file. What happens to it is different.
Take MyPresenter AI as the example. It reads the upload page by page and writes the narration for each page from what is actually on that page, rather than summarizing the file as a whole. Out of the same document it also builds a knowledge base — the set of things it is allowed to answer from later. Both are yours to edit before anything goes out.
What you get back is a link. The slides play with narration, and there's a chat bar beside them. Every page is open for questions, not just the ones you expected to be hard, so a student stuck on slide 12 asks on slide 12 and gets an answer without waiting for you. You send the link, or open it fullscreen on the classroom display.
You can also set the questions yourself rather than leaving it entirely to whoever shows up — write the ones you know are coming, attach them to the page they belong to, and answer them in your own words.
There is no MP4 at the end of this. The output is the link and nothing else — no downloadable video file, in any format, at any point. If a file is what you need, that is route one, and it is a different tool.
MyPresenter AI isn't the only tool that works this way; a few others in this category put a chat panel next to the playback instead of exporting a video. It's worth checking whether the one you're evaluating hands you a file, a link, or both, because the marketing copy here is nearly interchangeable and the output format is where the real difference lies.
Where the answers come from#
This is the question to ask about any tool in this category, and it's the one their pages tend to skate over.
There are three possible answers, and only one of them is safe for teaching:
From a general model. The AI answers from whatever it learned in training. Fluent, fast, and capable of confidently telling your students something that isn't in your syllabus and isn't true.
From the general model, "grounded" in your material. Better, but "grounded" hides a lot of variation. Ask what happens when the question isn't answered anywhere in your document.
Only from what you uploaded. The AI answers from your own file and nothing else, rather than from whatever it picked up in training.
MyPresenter AI takes the third route. You can also write reference answers ahead of time for the questions you know are coming, and flag certain questions as required, so students have to work through them before moving on.
For teaching, this isn't a nice-to-have. A wrong answer from a general model doesn't just confuse one student; it gets repeated back to you in an exam, by twelve people, and you won't know where it came from.
Where it runs out#
No file, as above. If you need an MP4 for an archive, a submission, or a colleague on a different platform, this route can't help you.
It needs a connection. The playback page runs in a browser and the AI runs in the cloud, so a student on a train through a tunnel is stuck — which wouldn't happen with a downloaded video on their phone.
And you're depending on a service. If you plan to teach this same module for the next five years, that's worth thinking about before you rebuild everything around it.
File or link: how to choose#
Pick video when the content is stable, the questions are predictable, offline access matters, someone needs a file they can keep, or the material has to travel outside any system you control.
Pick a presentation that answers when you teach the same thing repeatedly to a new group each time, when students need it to review before an exam rather than to learn something for the first time, when the same handful of questions keeps reaching you after class, or when you update the material often enough that re-rendering a video is its own chore.
Most teachers who do this end up on both routes. The settled modules become video. The parts students get stuck on become something they can ask questions of.
What students see on the other end walks through the same lesson from their side of the screen.
What neither one does#
Every page you'll find on this topic lists what its tool can do. None of them list what it can't. Here's the part that saves you an afternoon:
Neither route writes your lesson. Both start from material you already made. If you don't have one, you're in a different category of tool entirely — and that category has better options than either of these.
Neither route grades anything. No scoring, no rubrics, no gradebook. The required questions can tell you what students asked and where they hesitated. That tells you how the teaching landed. It doesn't tell you how the student did.
Neither route generates quizzes from scratch. You can attach questions to specific pages. The tool won't invent a summative assessment from your syllabus.
Both are picky about file formats. MyPresenter AI reads PDF, one presentation at a time. PowerPoint and Keynote have to be exported to PDF first. If you're planning a term's worth of material in an afternoon, there's no batch upload in the standard product — you'll be doing them one at a time, or asking about a custom setup. The supported upload formats page has the current list.
Neither replaces you in the room. Both handle the explanation you've given before. Neither reads the room, notices the confused silence, or decides on the spot to abandon the plan and do it differently.
Your first lecture, step by step#
Don't start with your best lecture. Start with the one that generates the most email.
Here's what the work actually looks like, using MyPresenter AI as the example. Setup runs about 3–10 minutes for one lecture.
1. Upload the PDF#
Export your slides to PDF first. PDF is the format the upload takes, up to 20MB — PowerPoint and Keynote have to be exported, and images and plain text aren't supported. One presentation at a time; there's no batch upload, so twenty lectures means twenty uploads.
2. Let it read the file#
It parses the document and lays every page out as a thumbnail, so you can see what it found before anything gets written. A forty-page lecture comes back as forty pages, not as a summary of the lecture.
3. Set how it should talk, then generate the narration#
Three settings, and they matter more than they look:
- How much it says per page. Brief, medium, or detailed. A dense methods slide and a section divider don't want the same amount of talking.
- Tone. A free-text prompt — informative, instructional, technical for a lecture; you'd write something else for a first-week orientation.
- Target language. You pick the language the narration comes out in. This is the setting that saves you a recording session per language.
Then it writes a script for each page, from what's on that page, and shows you the whole run as a list before you commit to anything.
4. Read the narration before anyone else does#
This is the step to slow down on. The draft will be roughly right and occasionally wrong in the specific way that matters — a term it hedges on, a step it skims. Fix it here, page by page, in the editor. Everything downstream inherits whatever you leave.
The same screen is where you pick who delivers it and what they sound like: a presenter, and a voice. Both are look-and-feel decisions, not content ones, and together they take a minute. There's also a translate action if you want a second language version of a lecture you've already tuned.
Each page has its own tabs for media, narration, and Q&A, so the work stays where the page is instead of in one long document.
5. Check what it's allowed to answer from#
The knowledge base is a thing you can open, read and edit — not a black box that got built once at upload. Fix a wrong entry, add the context the slide assumed, delete what shouldn't be said out loud. If the source changed enough, you can have it parse the document again and rebuild.
Whether the lesson ends up answering well or answering plausibly is decided here, and it's covered below.
6. Preview, then publish#
Preview plays it the way a student will see it, which is when you catch the things reading the script didn't reveal. Publishing gives you a link that anyone you send it to can open. Send it to students, put it in your LMS, or open it fullscreen on the classroom display and let it carry the part you've explained a hundred times while you take the questions.
Two things the steps won't prompt you for#
Before you share it, write the answers you already know are coming. You know them. They're the ones in your inbox every year at the same point in the term. Writing them down once is the difference between a lesson that answers well and a lesson that answers plausibly.
After you share it, read what students actually asked. That list is the most useful thing to come out of the exercise. It tells you which slide is doing less work than you thought, and it's the one output neither route gives you unless you go looking for it.
Then decide whether that lecture wanted a file or a page. You'll know after one term, and you'll be wrong about at least one of them, which is fine. The slides are still yours either way.
How this plays out across a whole course covers what changes when it's not one lecture but a term. The pattern shows up outside teaching too: the same pattern outside the classroom looks at it from a corporate training angle, where the audience watches once instead of coming back.