Hear them before you build

Hear our English voices. Native speakers in eight languages, 112 in German alone.

Start for freeSee pricing

  • Noah Tremblay

    Quick and clear.

  • Olivia Taylor

    Composed and warm.

  • Emma Gallagher

    Easygoing and clear.

  • Samantha Ferris

    Bright and helpful.

  • Gustav

    Deep and steady.

Native voices in eight languages

Every sample is spoken by a native voice. Tap to hear it.

Try it free

Try a voice on your own words

Up to 300 characters, read aloud by a native voice. No sign-up needed.

159/300

Make it yours

Recording
Your voice

Clone a voice from under a minute of audio

Upload clean audio of one speaker, then use it like any voice.

Cloning docs
Postgres
post-gres
Kubernetes
koo-ber-net-eez
API
A P I

Teach it your brand names

A dictionary fixes how a word is spoken in every request.

Dictionary docs

One ID, every mode

The same voice_id works for generation, streaming and streaming sessions.

Voices docs
From docs/features/voices.mdxpython
audio = client.tts.generate(    text="Hello with a specific voice!",    model_id="kugel-3.5",    voice_id=1071,)

Questions about voices

How many voices does KugelAudio offer?

The voices here are showcases with native samples; the public catalog holds 112 German voices alone. The voices API lists the whole catalog and filters it by language.

Which languages are the voices recorded in?

The languages section shows every native sample, each spoken by a voice of that language. The model speaks 39 languages, so a voice can also read a language it was not recorded in.

Why does every language have its own voice_id?

Each language is its own voice with its own voice_id, not one voice given an accent. Pass the voice_id of the voice you want to hear.

How much audio do I need to clone a voice?

Under 1 minute of clean speech from one speaker, as WAV, MP3, OGG, M4A or FLAC. Every clone request confirms you are the speaker or have their permission.

Can I control how brand names are pronounced?

Yes: a pronunciation dictionary maps a word to a replacement or IPA and applies to every request of its project. For a single request, write IPA inline between slashes.

Is generated audio marked as AI-generated?

Yes. Every synthesis response, including speech from a cloned voice, carries an in-band watermark and an HTTP disclosure header.

Pick a voice and start building

Create a free account for longer texts, every voice and the API.

Start for freeSee pricing