3 things to know about AI and audio

How I used AI to make a free promo video and got a pub offer

As someone with an Australian accent who wrote a book set in London, Zanzibar and Israel, I quickly ruled out narrating anything myself.

While I would only use a proper human voice for a full audiobook (see my post about the Uncanny Valley), here are 3 things I learned about AI for audio:

1. It’s easier than you think, and the top solutions have a free version. Upload your text, pick a voice, and you have a sample almost immediately. I used an AI voiceover for my book promo on Canva, which auto-generates subtitles too (this is fine for short projects like a promo video). The hardest part is picking the right voice (gender, accent, age, tone). There are a lot of them, and it takes a few tries. But once you find one that fits, it moves fast.

2. It will butcher anything that isn’t an English word. For example, my book has a smattering of Yiddish words. Remember that AI narration reads what’s on the page, and if what’s on the page is “chutzpah” or “Shabbat”, it will do its best and its best will not be good. The fix is simple-ish. Before you upload, go through your manuscript and rewrite the tricky words phonetically in a separate script version e.g. “chutzpah” as “HOOTS-pah”, “Shabbat” as “sha-BAT.” The AI reads your script, not the word, so you’re just telling it what to hear. Some companies have a pronunciation dictionary feature if you have a lot of them and want a more permanent solution. It’s fiddly but fixable, don’t let it put you off.

3. Go for a quick win. Using AI for your whole book may seem daunting. There are smaller quick wins that AI can give e.g. a first chapter sample. It’s a great way for loyal readers and/or new audiences to get a taste of your book. A 10-minute audio sample does what a cover and a blurb can’t, it puts a reader inside the book. Put that perfect opening chapter you wrote as a hook for agents or publishers to good use.

And one bonus thought on AI narration more broadly: don’t think about it only in terms of speed and efficiency. Think about personalization which is a true unlock of AI. Some readers might prefer a woman’s voice, others want a British accent, and some might want both or neither. You don’t have to do the whole book to test this. A first chapter in two or three different voices costs almost nothing and can be a great way to engage readers.

If you’ve read this far, I popped my Canva AI-narrated promo video on Instagram and got approached by a publisher, leading to an offer.

Magic can happen, good luck!