How to make a voice-over for a presentation with Text to Speech

A typical case: you have a 10-slide pitch or a lesson and need a narration for each slide, so the deck can be sent as a video or played without you. Recording it yourself takes many attempts, and any change means recording again. With Text to Speech you edit the text and generate the audio again.

Which tool to use and why

Use Text to Speech: it turns a written script into speech in a chosen voice. If the narration should sound like you, clone your voice first and pick it on the My voices tab: see what a custom voice is. Voice Changer doesn't fit here: it changes the voice in a recording you already have and doesn't work with text.

Step by step on an example

Example: a 10-slide product presentation in English, about 150 words of script per slide.

1. Write the script slide by slide in a notes app. Write numbers, abbreviations and names the way they should be read, for example "twenty twenty-six" instead of "2026" if you want it said that way.

2. Open Text to Speech on the Tools tab and paste the text for slide 1 with Paste. Tap Continue.

3. Choose one voice for the whole deck. For a business presentation try MiniMax Expressive Narrator or an OpenAI voice such as Onyx or Coral with the News preset. Tap Continue.

4. Tap Generate and listen to the result. If the pace or tone is off, adjust the speed for MiniMax or the instruction for OpenAI and tap Generate again.

5. When slide 1 sounds right, save the file: the button next to the player shares it in the app or downloads it on the web.

6. Go back to the Text tab, replace the text with slide 2 and generate again. The voice and its settings stay selected.

7. Insert each file as audio on its slide in PowerPoint, Keynote or Google Slides.

Save each file right after it's made. Without Premium, History keeps only the last 3 results, so after slide 4 the file for slide 1 is gone from History.

How to make the narration sound natural

• Voice slide by slide. Short parts are easier to fix and to place on the slides. One run takes up to 10,000 characters.

• For a pause before an important point, add <#1#> with a MiniMax voice or an ellipsis with an ElevenLabs voice.

• If a word is stressed wrongly, write the stressed part in capitals: REcord or reCORD.

• For a script in your own voice, record the voice sample in the language of the presentation.

More about tags and settings: how Text to Speech works and instructions for OpenAI voices.

How much it costs

For the example, 10 slides of about 150 words each, about 9,000 characters in total:

Every "Generate again" is charged as a new run, so test the voice on one slide before voicing the rest. Prices for all tools: token pricing for each tool.