How to Make an Audiobook With AI: Narrate a Whole Book on Your Own PC

Voice 13:36 2026-11-04
How to Make an Audiobook With AI: Narrate a Whole Book on Your Own PC

Watch on YouTube

AI-generated audio with Demodokos.

Learn how to make an audiobook with AI in Demodokos Foundry and narrate a whole book on your own Windows PC. Paste in the text, give every character a voice, direct the emotions and speaking styles, add music and generate everything in one go. Then an AI agent does the same job from one typed sentence. The video narrates the first three chapters of Dracula, which took 44 minutes on one PC. Calculated from that run, the whole book comes to about 7 hours of generation for about 18 hours of listening.

Written Tutorial

Prefer to read? The same walkthrough, step by step.

What you will make

A whole book as an audiobook, with a voice for every character, emotions, voice effects and music, all made on your own PC. The video works through the first three chapters of Dracula in five stages, by hand. Then an AI agent does the same job from one typed sentence.

1. The text goes in

Open Foundry, go to Text to Speech, paste in your book and press Process Text. Foundry's own small AI runs on your computer, reads the book and finds out who is in it. Then press Narrate Text, and every line goes to the right character. The wall of text is now a script, with a name on every part.

2. The characters get voices

Each name gets its voice in the cast list. Press Create New Voice, set the gender, age, pitch, pace and tone, and Foundry offers you voices that never existed before. Keep the one you like, and it stays the same from the first page to the last. A ready voice works too, and you can clone your own voice from a short recording and read the book yourself.

Foundry has more than 30 voice effects for the characters that need one. In the video, Dracula gets a deeper voice with an effect on top.

3. Direct the performance

Every line can have its own emotion or speaking style: whispering, pleading, nervous, angry, tender, grieving, even drunk. There are up to 38, and you choose how strong each one is. You won't do this on every line of a long book. Set how a character normally talks once, then step in for the big moments.

The way you type directs the voice too:

  • Spelling. Spell a line the way a frightened man would say it, stumbling over the words, and the voice stumbles with him.
  • Dots. A few dots make a pause.
  • Capital letters. They make a shout.
  • Round brackets. A word in round brackets becomes a real sound, like a booming laugh.

When a voice meets a word it has never seen, add that word once to Hot Words, spelled the way it sounds, and it comes out right everywhere in your script.

4. Add music

Foundry makes the music too. Describe it in plain words, like slow strings and a low organ, and you get an original piece. Then tell your script where it starts, where it stops and how loud it is. Give a chapter its own theme, or lay a quiet atmosphere under a scene.

5. Make it in one go

Press Generate Speech, and your computer makes the whole script by itself, line after line. Listen to it, and if a line bothers you, make that one line again, as often as you like. Press Export, and your chapter is a finished audio file.

One book, ten languages

The script is still in Foundry, so it can go further. Open Dubbing, choose a language, and Foundry translates the script and your cast speaks it. Foundry speaks ten languages, so one book can reach a lot more readers.

Hand the whole job to your AI agent

An AI agent is a program you talk to in plain English, and it does the work on your computer. You don't need to code. Foundry comes with written instructions for agents, inside your Foundry folder: documentation\agent_skills.

Put your book in a folder, open your agent and type one sentence:

The sentence from the video

Narrate the book in this folder, with a voice for every character and music for every chapter.

The agent splits the book into scripts, sets up the cast, the emotions and the music, and presses the button for every chapter. It checks its own work too. Foundry's speech recognition writes down what each line really says, a small script compares that with the book, and a line that doesn't match gets made again.

The best part stays with you. You press play and hear your own book, and if you want one moment told differently, you tell the agent and it makes that one line again.

The real numbers

  • Measured. Three chapters of Dracula took 44 minutes on one PC: 1 hour 53 minutes of audio in 296 parts.
  • Calculated from those chapters by word count. The whole book is about 18 hours of listening, about 7 hours to make, and about 9.5 hours with the word check and the music. Press start after dinner, and it is done before sunrise.

Which plan you need

Foundry is one flat price with no credits and no fee per character, so one book or a whole library costs the same.

On the Creator plan one script holds about 13 minutes and up to 5 voices, so you go through a book in small pieces, and you can make as many as you like. The Professional plan holds about 7 hours and up to 16 voices in one script, dubs with your whole cast, and lets your own AI agent run the whole job. Your agent has its own subscription, which is not part of Foundry's price.

Try it on your own first chapter with the 7-day free trial. See the plans.

What you need

Windows 10 or 11 and an NVIDIA graphics card with 6 GB or more (12 GB recommended). AMD cards with 8 GB are supported as experimental.

Your book stays yours

Foundry keeps your book, your voices and your audio on your own computer. Your book is never sent to our servers, and nobody at Demodokos reads or reviews what you make.