Record a short sample or upload audio. The service builds a private voice model, and in the studio you save delivery presets for ads, podcasts, audiobooks and IVR.
Read the suggested text aloud — about a minute. You can also upload an existing recording (WAV/MP3).
Confirm that this is your voice or that you have permission. Without this, a custom voice cannot be created.
Within a couple of minutes the voice appears in “My voices”. Create delivery presets and select them in the editor.
A delivery preset does not create a new model. It stores the instruction, speed, processing and volume on top of your base custom voice.
A fast, bright and confident preset for promos, announcements and short spots.
A conversational delivery with soft intonation for episodes, clips, short videos and educational explainers.
A calm narrator preset for long text, chapters, courses and meditative content.
Voice your videos in your own voice without re-recording every time.
Update lessons and add material in the same voice as the whole course.
A single recognisable company voice across all videos and notifications.
Creating a custom voice is included in the paid plans. The limit is the number of simultaneously active custom voices.
3
voices
5
voices
10
voices
20
voices
About one minute of clean speech is enough. The more varied the intonation in the sample, the more natural the result sounds.
Only if you have the rights. Creating a custom voice requires confirming consent. Uploading someone else’s voice without permission violates the service rules.
In 10 core languages with a quality guarantee: German, English, French, Spanish, Italian, Portuguese, Russian, Chinese, Japanese and Korean, plus 14 languages in beta via auto mode (all languages). You can record your voice in any of the 24 languages offered in the wizard — for beta languages synthesis runs via Auto. The best result usually comes from a language close to the sample.
It is a saved set of settings on top of your base voice: instruction, speed, processing and volume. For example, the same voice can be saved as “narrator”, “podcast” and “audiobook”.
The sample and model are stored on servers in EU, are never used to train shared models and are deleted together with the voice.
Available on Starter and above. Record a sample right in your browser.
Create your voice →