Cast an Audiobook Automatically with AI

Casting does the multi-voice setup for you. An AI service reads your manuscript, works out which character speaks each line, and lists every character it found. You then pick or design a voice for each one. It can also tag the mood of each part so the narration is read with feeling. Casting works on a multi-voice novel and on a book with a single narrator.

Before You Start

Casting uses a language model, which works a little differently on each version of the app:

• VCP Cloud has an AI service built in. Runs use tokens from your account, and there is nothing to set up.
• The desktop app uses the AI provider you connect in Settings, under AI Models, in the Audiobooks section. You pay your provider directly for what the run uses. See Connect an LLM to add a key.

If no AI service is connected on desktop, the Find speakers dialog shows the connection form so you can add one right there. You can also click Open settings.

Start Casting

There are three ways to start:

• Right after import. When a new book opens, the Your book is ready dialog asks what to run. Tick Find speakers, Tag emotions, or both, then click Start. Click Skip to set up voices by hand. The app remembers your choice for the next book. On desktop, this dialog appears only once an AI service is connected.
• From the Voices panel. In the left column of the editor, click Find speakers or Tag emotions above your voice list.
• For one segment. Hover a segment and click Find speakers to cast just that segment.

Before a run starts, the dialog shows Est. tokens and, on desktop, Est. cost. Actual use is often higher, and reasoning models use more tokens.

Find Speakers

Find speakers reads the book in parts and marks which character speaks each line. Nicknames and other names for the same person are joined into one character, and first-person narration is linked to the narrator.

While it runs, the dialog shows which part it is on, for example Part 3 of 12. Closing the dialog keeps the run going, and a bar across the top of the editor shows its progress. Click View progress to open the dialog again, or Stop to end the run early. Speakers found before you stop are saved.

Character voices need a model that can clone. If your book uses one that cannot, the dialog says so and the book switches to OmniVoice when the run starts.

If a few parts of the book could not be read, the dialog says how many. Dialogue in those parts stays with the narrator. Run Find speakers again to retry them.

Pick a Voice for Each Character

When Find speakers finishes, the Pick a voice for each character dialog opens. You can open it again any time with Pick voices in the Voices panel.

Each character is listed with its color, and with its gender, age, and the number of lines it speaks when those are known. The narrator is listed first. For each character you can:

• Click Pick a voice to choose any voice from your library or a built-in speaker.
• Click Design a voice to create a new voice from the traits the AI picked for that character.

Click Design 4 voices (the number matches the characters left) to design a voice for every character that still needs one. On VCP Cloud, the estimated token cost shows next to the button. Characters waiting their turn show Waiting, and Stop designing ends the run.

A designed voice is named after its character, for example Elena (designed). Click the play button to hear it, or Change to pick another voice. The description used to design it shows below. Click it to edit the text, then click Regenerate to design the voice again.

Click Save voices when you are done. Designed voices are saved to your library, so you can use them in other books too.

Characters Without a Voice

A character that has no voice yet shows Needs a voice in the Voices panel. A bar across the top of the editor counts them, for example 3 characters need a voice, with a Pick voices button. You cannot generate until every character has a voice.

Character Voice Design Model

Designed voices are made with Qwen3 or OmniVoice. When both are available, pick one under Character voice design model on the New project screen before you import the book.

Tag Emotions

Tag emotions reads your book and marks the mood of each part, such as happy, sad, or angry, so the narration is read with feeling. It works on its own for a single-narrator book, or together with Find speakers.

Emotions need Qwen3. If your book uses another model, the button reads Switch & Tag emotions and the book switches to Qwen3 when the run starts.

After tagging, you can change the mood of any segment under Segment emotion in the right panel, or of a highlighted span in the Selection panel. Pick an emotion and an Intensity from Very mild to Extreme. Emotions apply to cloned and designed voices.

Running Tag emotions again replaces every emotion in the book, including ones you set by hand.

Emotion Engine

Two engines can read a line with emotion:

• Steering applies the emotion straight to your cloned voice. It is faster and stays closer to your recording.
• Voice conversion acts the emotion with a different voice, then converts it to yours. It is slower, and the conversion can soften the emotion.

Each engine offers a different set of emotions. Set your default in Settings, under Advanced, with Emotion Engine. To change it for one book, open Project Settings and use Emotion engine under Generation.

If you switch engines after tagging, a bar at the top of the editor reads Emotions out of date. Click Tag emotions again to tag the book with the new set, or Keep existing.

Pause Between Speakers

When the voice changes inside a segment, the app adds a short pause so speakers do not run into each other. Set the length under Pause between speakers in the Advanced section of Project Settings. The default is 300 ms, and Off turns it off. Segments read by one voice are not affected.

Create a Dramatized Version

On the desktop app, Create dramatized version under the project name in the left column makes a new project with your book rewritten as a script, with speaker lines and emotions. The original project stays unchanged.

1. Click Create dramatized version.
2. Name the new project.
3. Turn on Fix mistakes with a second pass if you want the AI to check its work. This uses more tokens.
4. Click Create project.

The new project opens when it is ready. The Reader is not available in a dramatized project.

Tokens Used

If you stop a run or it cannot finish, the dialog shows how many tokens were used. On desktop it also shows what they cost, for example 48,200 tokens used, cost $0.03.

Deleting a Book

When you delete a book that designed voices, the confirmation includes a box such as Also delete the 3 voices this book designed. It is ticked by default. Untick it to keep the voices in your library.

Next Steps

• Voice Assignment to adjust voices by hand after casting
• Segments to change the speaker or emotion of a single line
• Use Multiple Voices in an Audiobook for a full multi-voice walkthrough