AI voice for Instagram Reels and YouTube Shorts: write a 30-second script that works

A short video gives you about two seconds to earn the next ten. That makes the voiceover the hardest part of a Reel or a Short to get right: there is no room for a slow start, a long sentence or a word nobody catches. This guide is about fitting a script into the time you have, and it starts with a number most people guess at: how many words actually fit.

How much script fits in a short video

At a normal speaking speed, a voice gets through roughly 900 characters a minute. That gives you a budget before you write a single line:

These are rough figures: speed, pauses and the language all move them a little. The useful part is the order of magnitude. If your 30-second script is 150 words, no voice setting will save it. Cut the script.

One number here is convenient. Without logging in, this site converts up to 500 characters at a time, which is almost exactly one 30-second clip. You can make a Reel voiceover without creating an account.

Write the hook first

The first line decides whether anyone hears the second. Write it last if you like, but give it the most time.

• Eight words or fewer. A hook the viewer has to wait for is not a hook.
• Start with them, not you. “Do you charge your phone overnight?” beats “In this video I will talk about batteries.”
• Promise something specific. “Three things” or “thirty seconds” tells the viewer how long they are signing up for.
• No greeting. “Hello friends, welcome back” costs two seconds you do not have.

One idea, three beats, one ask

A structure that fits in thirty seconds almost every time:

1. Hook: one line, a question or a surprising fact.
2. Three short points: one sentence each. Say “first”, “second”, “third” out loud; it helps the listener keep count.
3. One ask: follow, save, or comment. Pick one. Asking for all three gets you none.

Voice and speed for short video

Short video is faster than long video, and the voice should be too.

• Speed: start at +10% and try up to +20%. Past that, words tend to blur together, so listen on your phone speaker, not headphones.
• Pitch: leave it alone at first. A small lift can add energy; a big one sounds like a cartoon.
• Voice: the two free Hindi voices, Madhur and Swara, are the place to start. Pick one and keep it across your clips so people recognise your videos by sound.

If you want a read with more character, the premium voices give you more control. With the Gemini AI voices you can describe the delivery in words, such as “excited, like a cricket commentator”. With ElevenLabs voices, the Stability setting controls how varied the read is: lower is livelier, higher is steadier. New accounts get 5,000 credits to experiment with; the pricing page has the details.

Add the voice to your clip

1. Download the MP3 from the converter.
2. Open your video editor, such as InShot or VN, and import your clips and the MP3.
3. Put the voice on the timeline first. Then trim each clip so the picture changes when the voice moves to a new point.
4. Add music under the voice at a low volume. In a short video the voice should win every time.
5. Add on-screen captions for the key words. Plenty of people scroll with the sound off.
6. Export in vertical 9:16 and upload.

Common mistakes to avoid

• Too many words. The script was written to be read, not said. Cut every sentence that does not add a new fact.
• Long sentences. In a short video, a sentence with two commas is usually two sentences.
• Reading out a link. Nobody types a URL from a Reel. Say “link in bio” or put it on screen.
• Numbers as digits. “20-80%” may be read in a way you did not expect. Write “बीस से अस्सी प्रतिशत”.
• A different voice in every clip. It makes a page feel like a collection of reposts.

Making longer videos as well? The YouTube voiceover guide covers script writing in more depth, including what YouTube's monetisation rules say about AI voices.