How to add narration to PowerPoint: four ways, compared honestly
Record inside PowerPoint, screen-record, outsource, or generate it. What each actually costs in setup and — the part nobody budgets for — in re-recording when the deck changes.
Adding narration to a deck is easy. The part that decides whether it was worth doing is what happens three months later, when slide 14 has a new number in it.
That is the axis these four approaches really differ on, and it is the one that never appears in comparisons. Setup cost gets all the attention; update cost is what you actually pay.
Option 1: record narration inside PowerPoint
PowerPoint has had this since 2013. Open Slide Show → Record, choose whether to record from the current slide or the beginning, and it captures your voice and any ink annotations per slide, with timings saved against the slide.
Where it works well. You already have the deck and a microphone, and the narration is genuinely yours. Timing stays attached to the slide, so editing the slide later does not necessarily destroy the audio. For a one-off internal recording this is often the right answer, and it is free.
Where it stops scaling. Recording is linear — you record the deck in order, and a stumble on slide 19 means re-recording slide 19 (or the whole run, depending on how you are working). Twenty slides of clean narration realistically takes several takes and an hour. There is also no script: what you say is whatever you say, which is fine for an expert and inconsistent across a team of twelve trainers.
The export path is also narrower than people expect. Narration exports cleanly to a video file, but the result is a recording of a slideshow, not a video with chapters — viewers scrub, they do not jump by slide.
Option 2: screen-record a live presentation
Present the deck and capture the screen, optionally with a camera window in the corner. Tools here are commodity: OBS, Loom, PowerPoint’s own export.
Where it works well. It is the fastest route from “I know this material” to “there is a video”. You get your own delivery, your own emphasis, and no synthetic artefacts. For thought leadership, internal comms, and anything where the presenter’s personality is the point, this is hard to beat.
Where it stops scaling. Everything is baked in. Change a chart and you re-record the whole segment, and you will not match the previous take’s pacing, so the replacement clip visibly differs from its neighbours. Audio quality depends on the room. And there is no transcript unless you add one separately — which matters more than it sounds, because a video without captions or a transcript is unusable for a chunk of any large audience.
Option 3: outsource the voiceover
Send the script to a voice actor, get back clean audio files, drop one per slide.
Where it works well. Quality ceiling is highest here. A professional reads technical material correctly, including the acronyms and product names that trip up both amateur recordings and text-to-speech. If the video is customer-facing and the brand is careful, this is the safe choice.
Where it stops scaling. Turnaround is days, not minutes. Cost is per project, so iterating on the script is a new invoice. And it inherits the same rebaking problem: the audio is a recording, so a change to slide 14 means a new recording of slide 14, plus a fresh edit pass.
Option 4: generate the narration
Either text-to-speech reading a script, or a full presenter video where a digital presenter speaks each slide’s script with the mouth movement generated from the audio.
Where it works well. Regeneration. Because the narration is produced from text, changing slide 14 means changing the text and re-rendering — the other 39 slides come out identical, with the same voice, pacing and delivery. For training libraries, compliance courses and anything that gets updated on a schedule, that property is worth more than any single video. The same reason makes other languages cheap: the script is text, so a Spanish version is a translation and a re-render, not a new recording session.
Where it struggles, honestly. Three failure modes are real and worth planning around:
- Pronunciation. Uncommon proper nouns, product names and technical terms get mangled. Every deployment we have seen needed a pass over the script for this. Listen to the result before you publish it.
- Script drift. A generated script is inferred from your slide, and inference is confident even when wrong. It can assert something your slide does not support. This is why script review before rendering is not optional — treat the draft as a starting point.
- Throughput. Rendering is GPU work and does not parallelise for free. A 40-slide deck takes a meaningful amount of wall-clock time, and if the queue is busy you wait behind it. Fine when you are converting a library overnight; bad if you promised someone a video in ten minutes.
The comparison that matters
| Setup for 20 slides | Change one slide | Consistency across presenters | Second language | Skill needed | |
|---|---|---|---|---|---|
| Record in PowerPoint | ~1 hr | Re-record that slide | Varies by person | Re-record everything | Low |
| Screen-record | ~30 min | Re-record the segment | Varies by person | Re-record everything | Low |
| Outsourced voiceover | Days | New recording + edit | High, if same artist | New recording + edit | None |
| Generated | Minutes | Re-render one deck | Identical by construction | Translate + re-render | Low |
Read the “change one slide” column first. If your deck will be updated — and training decks always are — the first three rows charge you the full cost again, every time.
So which one should you pick
- Record it yourself if the video is a one-off, your delivery is the value, and the deck will not change.
- Screen-record if you need something today and it will be superseded rather than maintained.
- Outsource if it is customer-facing, the script is stable, and pronunciation has to be right the first time.
- Generate if the deck is one of many, it gets updated, or you need it in more than one language.
Most teams end up with a mix: recorded video where a person is the point, generated video where the material is the point and there is a lot of it.
Trying the generated route
If updating cost is the thing biting you, the cheapest test is your own deck rather than a template — a two-slide preview tells you nothing about how your charts survive, and “does my content come through intact” is the actual question. Signing up includes enough credit to convert a six-page extract of a real deck, which is short enough to check quickly and long enough to see whether the output is publishable.