How to turn a 60-minute lecture into 6 reels: a workflow for educators and experts

A university lecture runs 60 to 90 minutes. Almost none of it reaches social media, and when a piece does, it's usually the static tripod recording dumped on YouTube with no edit. That's a waste. In our experience an expert lands a strong hook every seven to twelve minutes of talking, so there are several ready-made hooks in every hour. The hooks are already there. What's missing is a way to get them out.

This is a concrete workflow, tested on three kinds of source: a university lecture, a corporate masterclass, and an expert webinar. From raw upload to six publish-ready reels takes about 90 minutes of your own time, polish included. That's time at the keyboard, not machine time. Do the same job without AI tools and you're looking at six to ten hours, depending on how handy you are with an editor.

Why lectures are the best source almost nobody uses

Educators and experts hardly ever make reels, which is strange, because they pack more usable content into an hour than anyone. A lecture already has a structure: setup, example, insight, transition, and around again. Each transition can become a standalone reel of thirty to sixty seconds, as long as you cut on the boundaries of a thought instead of on a stopwatch.

The real reason behind "I don't make reels" is how much work it looks like. An hour of lecture is roughly 9,000 words of transcript (the Wistia 2025 average puts conversational speech around 150 words a minute), and scrubbing through it by hand for the good bits takes two or three hours, on top of three to five hours of editing and branding. If you post one reel a week, that works out to about six hours per reel. No expert without a producer is going to keep that up.

AI cutting changes that math, not by replacing the editor but by taking the finding-the-moments part off your plate. In Lotima the software transcribes the lecture, reads how it's built, and pulls out the strongest moments: a hook and the key beats worth cutting. You ask for a reel from the best one, look over the edit plan the agent suggests, then come back and ask for the next.

Four steps from lecture to six reels

Step 1. Filming (10 minutes to set up, then 60 to 90 to record)

The costliest mistake is recording the lecture the usual way, with nothing set up for cutting it later. Three things you can't skip:

A stationary front-view camera, 1080p or better, 30 fps. An iPhone on a tripod two or three meters back is enough. Eye level isn't critical, since the vertical crop happens later, but keep the lens no lower than the speaker's jaw, or reframing shoves the face into the bottom third.
Clean audio. A lavalier mic on the speaker (a Rode Wireless Go II runs about $200, and the clip-on lav is the bare minimum for usable sound). A camera mic three meters away leaves a quarter to nearly half the words unreadable to AI transcription, and everything downstream falls apart with it.
A one-sentence payoff every seven to ten minutes. This is just something to tell the speaker beforehand: try to land the conclusion of each section as one self-contained sentence. It lifts automatic moment-selection a lot, because a strong sentence at the end of a stretch is exactly the signal a narrative-aware algorithm reads as "this thought is complete".

Step 2. Upload (3 to 5 minutes)

Upload the file to your auto-editing tool. What matters:

Size: a 60-minute lecture in 1080p usually weighs 3 to 6 GB. A steady 50-plus Mbps connection sends it up in four to eight minutes. Lotima takes sources up to 8 GB, or imports straight from a YouTube link or cloud storage (Google Drive, Dropbox, and others). On a slow connection, run the upload overnight.
Context: say what you want in plain words. The language is detected for you. What helps is telling the agent it's a lecture and what length you're after (thirty to sixty seconds by default, sometimes sixty to ninety for teaching content, depending on the platform).
Brand voice: before the first reel goes out, settle on a caption font, a main color, and a corner logo. In Lotima you can also show the agent a reel or two as a reference, and it picks up your format for the next render.

Step 3. Picking the moments (15 to 20 minutes)

Once the transcript and narrative analysis are done, the agent shows you the strongest moments: a hook and up to about ten key beats, each with a reason for why it would stand on its own as a clip. One pass makes one reel, so you pick the moment you want first and the rest wait for your next asks. This is the step that matters most. It decides whether anyone wants to watch the finished reels.

What to look for in the candidates:

Does it stand alone? Would you understand the moment from those sixty seconds with none of the lecture around it? If it only makes sense with the previous slide, skip it.
Is the hook strong? The first sentence is your bid for attention. "Today we'll talk about..." fails. "Most people think X, but the data shows Y" works.
Is there a payoff? Is there an aha in the clip, a concrete insight that makes watching to the end worth it? Descriptive stretches with no insight kill retention.
Is the tempo right? Too slow, one phrase every five seconds, drags in short form. Too fast, three phrases a second, is unreadable. The sweet spot for short form is 120 to 180 words a minute.

Expect about half the moments it surfaces to be worth a reel. The rest either echo their neighbors or fail one of the four tests above. Don't take it as the tool underperforming. Throwing out the extras is just part of editing, the same as it's always been.

Step 4. Render and branding (20 to 30 minutes)

Once you sign off on the edit plan, three things happen in a single render:

Reframing from 16:9 to 9:16 with face-aware framing that keeps the speaker composed in the vertical frame instead of blindly center-cropping. If the lecture has slides, ask for a split layout (speaker up top, slide below) or picture-in-picture. The montage catalog also covers cutouts, grids, and over-the-shoulder layouts, and the agent will tell you straight if your footage can't support the one you asked for.
Word-level captions, with a karaoke-style highlight if you want it. Font, size, and color come from your brand setup, and highlighting keywords is a one-line ask in chat.
Pacing: dead pauses trimmed automatically and the rhythm smoothed, with a mode to match the material (aggressive, balanced, or story). You can add music from the built-in catalog, but for lectures we'd leave it off, since it undercuts credibility.

Each reel renders in the background while you move on to the next one, and the captions come baked into the finished 1080×1920 MP4.

What makes a lecture moment actually good

In a podcast the energy comes from the back-and-forth: the question and answer, the guest's reaction. A lecture is different. Its energy rides on one of three shapes:

Counterintuitive claim plus evidence: "Everyone thinks X, but N studies show Y." The strongest shape for teaching content in short form. Usually two or three moments in a one-hour lecture follow exactly this pattern, and they should land near the top of your candidates.
Concrete example plus the general point: "Here's a case, one student did this. And it shows the broader principle Z." It works because it opens with something concrete, which is easy to follow, and closes on the abstraction, which is what sticks.
Three-step framework: "Here are three ways to solve this: first, second, third." Platforms love it, and three-step content reliably holds attention above the niche average. The catch: it has to fit in thirty to sixty seconds, no dragging on to a fourth and fifth step.

What doesn't work in short form: a long windup with no conclusion, historical asides (unless you happen to be Dan Carlin), open "let's think about this together" questions where you never give your own answer. All fine in a full lecture. In a clip they land as an unfinished thought.

How to brand academic reels

The default creator preset, bright colors and emoji in the captions, does not work for academic short form. That audience reads it instantly as infomarketer production values, and your credibility drops with it. What works:

A dark caption background that stays readable over both the light and dark parts of the frame. Text in white or a light cream.
A geometric sans-serif with no decoration: Inter, IBM Plex Sans, Source Sans 3. No handwritten fonts, no outlines.
Keyword highlight through weight, not color: a light bold (font-weight 700 over the body's 500) rather than a colored background. Easier on the eye.
A lower-third with the expert's name and credential, on screen in the first three seconds of every reel. It's an E-E-A-T signal for the viewer and for the platform's algorithms both.
A small corner logo or watermark, no shimmer. For an academic audience, a watermark covering a third of the screen reads as "selling a course".

Edutopia and the rest of the microlearning research all point the same way: visual restraint raises how credible you look, and that counts for most when the material is dense.

Why scoring a lecture is not like scoring a podcast

Most AI editors run the same moment-scoring model on podcasts and lectures alike. From an engineering seat that makes sense, since one model is less to maintain, but it costs you in the quality of what comes out. A podcast and a lecture are built differently:

In a podcast, the energy comes from the speakers trading off: line, reaction, punchline. A model trained on podcasts goes hunting for those interaction beats, the moments where the speaker changes and one of them peaks.
In a lecture, the energy comes from the logic: claim, evidence, insight. The model has to catch the shift in meaning, not the shift in sound. A dropped voice in a lecture usually means "I'm getting to the conclusion", and cutting there throws away the payoff.

That's why Lotima leans on the transcript's narrative instead of the loud spots in the audio. The analysis looks for shifts in meaning (claim, evidence, insight), and the agent tells you why it thinks a moment is complete, so a payoff that got clipped short is something you catch in the edit plan, before the render, not in the finished clip.

What this means in practice: if you work in mixed formats (a weekly talking-head, a monthly lecture, a quarterly interview), test a tool on a lecture before you commit to it. Run an hour through and watch where the cuts fall, on the boundaries of a thought or on the loud spots. You'll see the difference in a single test.

Where to go from here

What's above is the bare-minimum version. From here you can add thumbnails pulled from the transcript, A/B test the first three seconds across variants of the same reel, build B-roll libraries tagged by keyword. All of that is worth doing once the basic process runs reliably.

Until it does, don't pile more on. One hour of lecture into six reels a week is already a big step up from where most experts sit, which is zero or one reel a week.

Blog