Create an AI voice

Last updated: May 13, 2026

This article covers every method for creating a custom voice on Wondercraft. Use a custom voice when the library doesn't have the exact sound you need.

Overview of creation methods

Wondercraft gives you five ways to create a voice. Each lives in a different part of the editor, depending on whether you're starting from scratch or building on an existing voice.

Method

What it does

Where to find it

Clone

Replicates a real person's voice from an audio sample (90% accuracy)

Voice tab > Add your own voice > Clone

Import PVC

Imports a Professional Voice Clone created outside Wondercraft (99% accuracy in terms of emotion, tone, pace, accent, etc; needs ~40 minutes of clean audio)

Voice tab > Add your own voice > Import PVC

Design/Prompt

Builds a voice from scratch using written instructions

Voice tab > Add your own voice > Design

Select an audio clip > Create new > Prompt

Speak like

Transfers the delivery style of one voice onto another

Select an audio clip > Create new > Speak like

Clone a voice

  1. Open the Voice tab in the left panel.

    Screenshot 2026-04-16 at 13.33.58.png

  2. Click Add your own voice at the top of the voice list.

  3. Select Clone.

  4. Upload an audio file or record yourself directly. Audio samples can be up to 5 minutes long and 50 MB. See below for audio recording best practices.

    Screenshot 2026-04-16 at 13.35.05.png

  5. Follow the prompts to name your voice and confirm.

You can find voice clones under the Voice tab.

Screenshot 2026-04-16 at 13.37.28.png

Import a Professional Voice Clone (PVC)

If you want a high quality voice clone that captures your tone, emotion, pace and accent perfectly, you should consider getting a Professional Voice Clone (PVC).

See 📄 Create a Professional Voice Clone (PVC) for details.

  1. Open the Voice tab in the left panel.

  2. Click Add your own voice.

    Screenshot 2026-04-16 at 13.33.58.png

  3. Select Import PVC.

  4. Follow the prompts to upload or connect your PVC file.

Screenshot 2026-04-16 at 13.44.44.png

Generate an AI voice by prompting

You can create an AI voice in the style, tone, accent and pace you want from scratch.

  1. Open the Voice tab in the left panel.

  2. Click Add your own voice.

    Screenshot 2026-04-16 at 13.33.58.png

  3. Select Design.

    Screenshot 2026-04-16 at 13.44.44.png

  4. Describe the voice you want: accent, gender, age range, tone, pace, and personality traits. For example: "A warm British female voice in her 40s, calm and authoritative, with a measured pace."

  5. Preview the generated voice and iterate on your description if needed.

  6. Save the voice to your library.

Create a variation of an AI voice (different accent, tone etc)

Speak like copies the intonation, pace, and accent of one voice and applies it to another. Use it when you want Voice A to deliver content the way Voice B sounds - transferring a specific delivery style or accent to a different base voice.

  1. Select a speech clip on the timeline.

  2. Under the Voice section, click Create new.

  3. Select Speak like.

    Screenshot 2026-04-16 at 14.23.07.png

  4. Choose a reference voice whose style you want to transfer.

  5. Preview the result and save.

    Screenshot 2026-04-16 at 14.23.30.png

Voice creation best practices

Recording the audio sample for your AI voice clone

For the AI to capture your voice accurately, follow the best practices when recording your audio clip used on Wondercraft.

Pace

Speak at the pace you want the AI to mimic, and keep it consistently fast/slow throughout the audio clip.

Tone

You need to be extremely expressive. When the AI clones your voice, it dampens the expressivity. So you need to overdo it so that after the cloning process, it is at a good level. So try to cover all the spectrum between enthusiastic and sad and please over do it, to a funny level.

Quality

Clear, no noise of any kind. Use the best microphone you have to minimize any audio interference.

Length

Aim for 30 seconds of audio

File Type

MP3, M4A, WAV, AAC, FLAC

File Size

No more than 10MB

Volume

Try to record loudly. Aim for 18db so that it is equalised with Wondercraft voices.

Script language

Read a script in the language that your clone will be speaking. If your clone will be speaking French, record the audio clip in French.

Improving the quality of your AI voice clone

If your cloned voice sounds robotic, flat, or otherwise unnatural, the issue is almost always with the source audio. The cloning model learns from what you give it, so higher-quality input produces a more realistic result.

To improve the quality of your voice clone, make sure:

  1. You have followed the best practices outlined in the "Recording the audio sample for your AI voice clone" section.

  2. You read the source script in an emotional tone. Make it even more dramatic/emotional than what is natural in day-to-day speech. If you read in a flat tone, the clone will sound flat too.

  3. Your source audio doesn't have any background noise or echo, otherwise the ambient sound in the recording will be baked into the voice clone.

  4. Your recording is between 30 seconds and 1 minute, otherwise the model would have less than what's needed to work with.