- Speaker notes travel inside the .pptx file. A PDF export drops them, so upload the .pptx.
- The AI writes narration from your notes and the slide; it does not invent facts that are not there.
- Slides without notes are narrated from the slide text, which usually sounds thinner.
- Two to four plain sentences per slide is the sweet spot. Bullet fragments work, full scripts work best.
- Every line stays editable before you generate audio, and nothing is charged until you do.
- Write notes for the learner, not for yourself: no 'remember to mention' reminders.
The short answer
Yes. Speaker notes are the closest thing most decks have to a script, and an AI narration tool can turn them into the voice-over for a video. Upload the .pptx, and the tool writes each slide's narration from that slide's notes and content. You review the script, fix anything that is off, and generate the audio.
The quality of the video depends on the quality of the notes more than on anything else. This guide covers how the notes are used, how to write notes that narrate well, and what to check before you render.
How AI narration uses your notes
When you upload a .pptx to Unmixr's PowerPoint to video workflow, each slide becomes one scene. For each scene, the AI looks at three things:
- The speaker notes, which carry what you meant to say.
- The slide text and the slide image, which carry what the learner will see.
- The rest of the deck, so the narration of slide 12 follows on from slide 11 instead of starting over.
From those, it writes narration that sounds like someone presenting the slide. It does not invent facts that are not in the notes or on the slide. If a slide has no notes, the narration is written from the slide text alone.
The script is a draft, not a final cut. Every line is editable before you generate audio, and nothing is charged until you do.
What different kinds of notes produce
| Your notes look like | What the narration sounds like | What to do |
|---|---|---|
| Two to four plain sentences | Accurate, natural, close to what you would say | Nothing; this is the target |
| Bullet fragments ("Q3 revenue; up 12%; churn flat") | Usually good, sometimes generic | Check numbers and emphasis in the script |
| A long, word-for-word script | Faithful but can run long | Trim in the script if a slide drags |
| Reminders to yourself ("mention the demo here!") | Confusing | Rewrite as what the learner should hear |
| No notes at all | Thin; describes the slide | Add notes to the slides that matter |
How to write speaker notes that narrate well
Write for the listener. Notes written as reminders to yourself ("don't forget the caveat") confuse the narration. Write what the learner should hear: "There is one exception: contractors follow a separate process."
Explain, do not recite. The learner can read the slide. Good notes say why the slide matters or what to notice: "The number to watch here is the drop in week three."
Keep it short. Two to four sentences per slide, roughly 30 to 80 words, is 15 to 40 seconds of speech. A slide with 300 words of notes becomes a two-minute monologue over a still image.
Spell out tricky terms. If an acronym should be read as letters or as a word, say so. You can also set a pronunciation rule later, which fixes the word on every slide at once.
Keep the order of the slide. If the slide lists three steps, put them in the same order in the notes, so what the learner hears matches what they see.
Step by step: from notes to narrated video
- Upload the .pptx, not a PDF. A PDF keeps the slides but drops the notes. Decks from Google Slides or Keynote need to be exported as .pptx for the same reason.
- Choose Narrated course video as the output and pick the narration language. It can differ from the language of your notes.
- Add a short guidance note if your audience needs it, such as "New hires in a hospital. Use plain language. Explain acronyms on first use."
- Choose the narrator, then Review in Scene Studio.
- Read the script next to each slide. Fix anything that changes your meaning, and check every name and number.
- Generate the audio, listen to the first few slides, and add pronunciation rules for any word that comes out wrong.
- Export the 720p MP4 with SRT and VTT captions and a transcript.
What to check before you render
- Numbers and dates. Read them aloud in your head: "12%" and "Q3 2026" should come out the way you would say them.
- Slide length. Narration that runs past what the slide shows feels slow; a busy slide with one sentence feels rushed. Split or trim.
- Hand-offs between slides. Phrases like "as we saw earlier" should refer to something the video actually showed.
- The first slide. It sets the tone for the whole video. Make sure it says what the learner will get from it.
When speaker notes are not enough
Some decks have no notes at all, or notes written for a live audience that do not work as a recording. In that case you have two options:
- Let the AI write from the slides, then edit. For simple decks this is often fine.
- Write the script directly in the editor. Scene Studio lets you replace any slide's narration with your own wording, which suits regulated content where every sentence is approved.
Next step
Pick a deck with good notes and turn it into a narrated video. A free account exports up to 3 minutes with a small watermark, so you can hear what your own notes sound like before you decide. For more on producing whole courses this way, see the AI training video maker and AI course narration.