Custom Voice

Build a recognizable, one-of-a-kind voice for your Text-to-Speech apps with your speaking data available. You can further fine-tune the voice outputs by adjusting a set of voice parameters.

Get started

New to Speech Services? Create a Speech resource

Create a voice as unique as your business

Do you need to give your voice agent a unique, recognizable brand voice? The Text-to-Speech voice customization feature makes it easy to create one-of-a-kind voice-enabled apps, with no expertise required. To customize your voice agent, simply record and upload training data, and Microsoft creates a unique voice font tuned to your recording. The system easily scales as you get more data to provide an even more natural voice.

When using either default or customized voices, you can further tailor your audio by controlling parameters such as the speed of speech, pitch, volume, additional pauses, and pronunciations.

Hear Custom Voice in action

Build a highly natural voice without a single line of code, starting from just a few minutes of audio.

BASIC

2 hours of speech data or less

STANDARD

3 hours of high-quality voice recordings

HIGH-QUALITY

8 hours of high-quality voice recordings

Jessa

Guy

Zira

Linda

Mark

Richard

Zira

Evan

David

Creating a custom voice model

1 Prepare training data and create a Speech resource before you start to train a Custom Voice.

2 Upload your data to the Custom Voice portal or through the Custom Voice API and check quality.

3 Use your data to train a custom model. Test the model with your script when it’s ready.

4 Deploy the voice model to get your custom API endpoint. Test the endpoint before you integrate it in your system.

5 Use the voice in your apps by using code samples from the endpoint.

Voice Tuning preview

Quickly refine high-quality voice output with an SSML-powered web tool

Customize and fine-tune spoken text to best suit your needs. Experiment with voice styles in real-time, tailor speech patterns, and quickly create high-quality voice output.

Custom Voice offers tools to easily tune standard, custom, and neural voices. Choose from a wide palette of SSML-supported voice attributes such as rate, pitch, volume, pronunciation, and breaks.

RATE

BREAKS

PRONUNCIATION

Customize output by <prosody rate="-50.00%"> slowing-down the speed rate.</prosody>

Add a break <break time="600ms"/> between words.

You can pronounce it ASAP or <sub alias="as soon as possible">ASAP</sub>.

Get started with Voice Tuning now

Start tuning

Custom Voice turnkey solutions

Premium services that take care of everything for you

State-of-the-art custom voices

The highest quality digital voices, fully-tested and fine-tuned, in 49 supported languages.

Custom Voice turnkey solutions now support Neural Text-to-Speech, the latest deep neural network technology to create digital voices nearly indistinguishable from human recordings.