Record a Narrated Treatment Walkthrough in PowerPoint and Export It as Video
Record a Narrated Treatment Walkthrough in PowerPoint and Export It as Video
You finish this job with two files, and they do different work.
The first is the presentation itself, editable, with the narration and slide timings still attached to the slides that carry them. The second is the movie: one file that plays on a machine that may not have PowerPoint, for someone who was not in the room when you first explained the treatment.
Most people produce one of those and call it done. Screen-record the deck in a capture tool and you get a movie with no editable source, so the next correction means a new performance of the whole thing. Record inside PowerPoint and skip the last step and you get a file nobody has watched, which is how a treatment goes out with a silent fourth slide and the wrong sentence sitting over the logo.
The deliverable is the reviewed movie. The editable source is what keeps the movie repairable.
Name the branch you're on, then put the source somewhere safe
The steps below follow Windows desktop PowerPoint. Mac PowerPoint, PowerPoint for the web and mobile versions place these controls differently, and advice that hops between branches produces menu items that don't exist on your screen. Microsoft documents the recording side in Record a slide show with narration and slide timings and the export side in Turn your presentation into a video. When my wording doesn't match a label in front of you, the version-specific pages are the ones to trust.
Before you record anything, save a copy of the deck and name it so you can recognise it later — something like LowTide_Treatment_Source. That copy is the working file. The exported movie gets its own name, and the two never share one. The reason is simple: you will want to change something in the source later, and you don't want to be guessing which file is which when a client is on the phone.
Decide now whether the walkthrough is audio alone or audio plus you on camera. A treatment is usually better served by the slides than by a face, and recording controls on several platforms will happily capture a webcam that happens to be connected. Find the camera setting before the recording starts, not by watching the result.
Give each slide one spoken job
You are not reading the deck aloud. The slides already carry the text, the diagrams and the tone reference; the narration carries what they can't say. If two slides are doing the same job for the viewer, the narration on one of them will contradict the other, and you'll discover it in preview, after both are recorded.
This matters for recording more than for writing. A slide with one job has one boundary, and a boundary is what lets you replace a single recording later without rerecording its neighbours. That's the whole argument for the next few sections: the size of your repair is set by how cleanly you divided the work up front.
For a three-slide treatment that might look like: slide one states the premise and the two things it resembles, slide two explains what changes for the protagonist over the season, slide three states the question still open and what you're asking for. If slide two resolves the question slide three is asking, you have two jobs fighting, and it will be audible.
Record a trial first, on a deck you don't care about
Build the smallest possible exercise: three slides, made this morning, content you would be happy to lose. This exists so that your first experiment is not the treatment.
Speak at the distance and volume you'd actually use. Play it back. Your voice should be the loudest thing in the recording; if the room is competing with you, fix the room or the microphone position before you go further. If the trial records a webcam, screenshare or pointer you didn't want, you've just found that out for four minutes of work instead of forty.
The trial also tells you how your version handles transitions, how long it takes to start recording once you press the button, and whether the controls you read about appear where you expected. Platform and version branches genuinely differ here, so a trial is faster than research.
Record with clear slide boundaries, then preview inside the presentation
Advance the slide, then start the sentence. Leave enough room for the visible change to register before the next explanation arrives, and don't let a sentence on slide three describe an image that was on slide two. Those two habits eliminate most of what you'll otherwise find in preview.
When the run is done, replay it inside the presentation rather than exporting to check. The documented recording workflow stores narration and timing on individual slides and gives you a preview of what you recorded, which makes this the cheap place to catch problems. Listen for the things that only exist in speech: a word swallowed at the start of a take, a pause where you were checking a note, the sentence that made sense when you could see the slide behind it and doesn't in playback.
Then do the thing most people skip: look at the stored state of the slides rather than assuming the last take you spoke is the one the file holds. Narration and timings belong to individual slides, and the state can differ from your memory of the session.
Replace the slide that failed, not the presentation
This is the point of the whole method. One slide is wrong — a name, a number, a sentence that resolves something it shouldn't — so you repair that slide.
Before you click, read the scope of whatever you're about to use. Commands that clear or rerecord a slide come in at least two sizes: the one that affects the slide you're looking at, and the one that affects every slide in the deck. The second one costs you the entire session, including the parts that worked. Microsoft's documentation describes both scopes and a preview step; your version may word them differently, so read the label rather than trusting muscle memory.
Where a version separates clearing narration from clearing timings, treat those as two separate operations, because they are. Replacing the spoken take on a slide does not necessarily replace the advance time the previous take set. That's worth pausing on, because it's the failure this article exists to prevent: a repaired sentence and a stale clock, sitting on the same slide.
Once the replacement is recorded, replay that slide and its two neighbours. What you're checking at the join is whether the picture changes before the new sentence finishes, or whether the next slide sits in silence after it has. Either one is the signature of a timing that no longer matches the audio.
Export the recorded version — then watch the file you exported
Save the source first, so the thing you're exporting from is the version you want to keep.
The export dialog includes a choice about whether recorded timings and narration are used. That single setting is why so many exports arrive silent and self-advancing. Choose the option that carries your narration and timing, and give the movie a distinct name with a version marker or date in it, so the one you watched is the one you can find again. If your version also offers a size or quality choice, choose for legible on-screen text over the smallest file — a treatment's diagrams are the argument.
Then play the movie from beginning to end, in a video player, not in PowerPoint. Working through the whole file is the only way to catch what sits between slides: the opening seconds before the first sentence, the join at the slide you repaired, transitions into and out of it, any inserted media, and the final hold at the end. An export is not evidence that your media, timings or transitions survived it. Watching the file is.
Two things a successful export does not prove, and both are worth a separate check rather than an assumption. Transcript or caption support, if you need it, is a separate question from whether the video plays. And the recipient's route — the link, the shared drive, the platform that re-encodes what you upload — is separate again. Confirm the movie on the path they'll actually use before you call it delivered.
A constructed walkthrough: three slides, one replacement
The case below is a construction, not a report. Nothing was recorded or exported to write it; it's laid out small enough to run yourself in about ten minutes, and it shows where the failure lives.
The deck is a treatment called Low Tide. Slide 1 states the premise and runs 22 seconds of narration. Slide 2 explains what changes for Mara across the season. Slide 3 poses the decision the treatment is asking the room to make, and runs 15 seconds. Slides 1 and 3 are fine.
Slide 2 is not. Its narration resolves the question that slide 3 presents as open — an internal contradiction the deck shouldn't have, and one that preview catches immediately. The fix is a rerecord of slide 2's narration, and slide 2 only.
Here's the catch. The first take on slide 2 was rushed, and it set a slide time of about 18 seconds. The replacement take is 31 seconds, because the corrected sentence takes longer to say. If that 18-second advance survives the narration replacement, the movie comes out 13 seconds shorter than the narration needs, and the mismatch shows up at exactly one place: the join between slides 2 and 3. If you watch that join and listen for the picture changing mid-sentence, or for slide 3 waiting in silence for a voice that has already stopped, you'll find it in a second.
The repair is not a rerecord. It's re-checking slide 2's timing — that slide's timing, not the deck's — and previewing the join again before exporting.
The claim this example is making is about slides 1 and 3. They kept their narration and their timings through the entire repair, because the recording state belongs to the individual slide and the repair was scoped to one. That's what makes a one-slide correction a one-slide correction rather than a re-performance of the treatment.
So keep the two files, and keep them named apart: LowTide_Treatment_Source, editable, carrying the narration and timings you can still change, and LowTide_Walkthrough_v3, the movie you watched from start to finish, with the corrected join and the final hold intact. The source is what lets you fix slide 2 again in a month. The movie is what you send.
Frequently asked questions
What two files should a narrated treatment walkthrough produce?
The job should leave two files that do different work. The first is the presentation itself, editable, with narration and slide timings still attached to the slides that carry them. The second is the movie: one file that plays on a machine that may not have PowerPoint, for someone who was not in the room. The deliverable is the reviewed movie, and the editable source is what keeps the movie repairable. Save a working copy with a recognisable name, and give the exported movie its own distinct name.
Why record a trial on a deck you don't care about?
Build the smallest possible exercise with three slides, content you would be happy to lose, so your first experiment is not the treatment. Speak at the distance and volume you would actually use and play it back. Your voice should be the loudest thing in the recording. The trial also reveals unwanted webcam, screenshare, or pointer capture, and it shows how your version handles transitions, how long recording takes to start, and whether the controls appear where expected. Platform and version branches differ, so a trial is faster than research.
How do you replace one failed slide without re-recording the whole deck?
Give each slide one spoken job, because a slide with one job has one boundary and can be replaced without rerecording its neighbours. Before clicking, read the scope of the command: some commands affect the slide you are looking at and some affect every slide in the deck. Where a version separates clearing narration from clearing timings, treat those as two operations. After the replacement is recorded, replay that slide and its two neighbours, checking whether the picture changes before the new sentence finishes or the next slide sits in silence after it has.
Why can a repaired narration still result in a movie shorter than the narration needs?
Replacing the spoken take on a slide does not necessarily replace the advance time the previous take set. In the constructed example, slide 2's first take set about 18 seconds, but the replacement is 31 seconds. If the 18-second advance survives, the movie comes out 13 seconds shorter than the narration needs, and the mismatch shows at the join between slides 2 and 3. The repair is not another rerecord; it is re-checking that slide's timing, not the deck's, and previewing the join again before exporting. Slides 1 and 3 keep their narration and timings because the recording state belongs to the individual slide.
What should you do after exporting the movie?
Save the source first so you are exporting the version you want to keep. In the export dialog, choose the option that carries your narration and timing, and give the movie a distinct name with a version marker or date. Then play the movie from beginning to end in a video player, not in PowerPoint. Check the opening seconds before the first sentence, the join at any repaired slide, transitions into and out of it, inserted media, and the final hold. A successful export does not prove transcript or caption support or that the recipient's link or platform route preserves the movie; confirm the path they will actually use.