Welcome to the Custom Neural Voice portal

Custom Neural Voice (CNV) lets you create a natural-sounding synthetic voice that is trained on human voice recordings. Your custom voice can adapt across languages and speaking styles, and is perfect for adding a one-of-a-kind voice to your text to speech solutions.Learn more about Custom Neural Voice.

Voice recordings and transcripts

Train by Custom Neural Voice

Synthetic voice for your brand

Hear how Custom Neural Voice works

Learn about the different options for creating a Custom Neural Voice

With Custom Neural Voice (CNV), you can create two types of projects, Lite and Pro.The following table summarizes key differences between the CNV Lite and CNV Pro project types.

Project type	Lite	Pro
Best for	Create a synthetic voice of your own in just under an hour; ideal for testing and evaluation	Design and create a best-in-class synthetic voice for your brand based on professionally recorded samples; ideal for real world scenarios
Voice quality	Moderate quality	Highly natural-soundingResembles the voice actors' accent and intonation
Voice samples (Default neutral style)	Lite, Male, English (UK)Trained with 40 voice samples recorded online Original voice recording Trained with CNV Lite	Pro, Male, English (UK)Trained with 300 professional studio recorded samples Original voice recording Trained with CNV Pro
Training requirements	Voice talentCreate a synthetic voice of yourself ScriptsRecording scripts provided on screen by Microsoft Recording settingsRecord your voice online in a CNV Lite project on your computer Required sample size20 to 50 recorded utterances Voice talent consentVoice talent consent recording required for deployment TrainingLess than one compute hour* to train * Compute hour is the unit used to calculate the cost of Custom Neural Voice trainings. Normally two computing tasks go in parallel when a voice is being trained.	Voice talentRecruit professional voice talents that meet your designed persona ScriptsWrite your own scripts to match your use case or use our sample scripts on GitHub Recording settingsRecord voice professionally with professional recording equipment Required sample size50 to 2000 recorded utterances depending on model versions Voice talent consentVoice talent consent recording required. Learn more Training20 to 40 compute hours* to train a single-style voice, 90 compute hours* to train a multi-style voice.
Speaking styles	Not available	Enable your voice to speak in multiple emotional styles without requiring additional style training data Create your own style with your own training data Pro, Female, English (United States)Trained with 500 professional studio recorded samples Neutral Angry Cheerful Excited Friendly Hopeful Sad Shouting Terrified Unfriendly Whispering Original voice recording
Cross-lingual adaption	Not available	Yes; Have your voice speak in additional languages with no extra training data needed. Pro, Female, English (United States)Trained with 500 professional studio recorded samples English (United States) French (France) German (Germany) Portuguese (Brazil) Chinese (Mandarin, Simplified) Korean (Korea) Japanese (Japan) Original voice recording
Supported languages for training data	13 languages	66 languages
Availability	Available to try out with your own voice with Azure Speech Standard (S0) resource	Access is limited in order to support Microsoft Responsible AI principles; Apply for full access to create a Pro voice. Learn more about access requirement here
Pricing	Per unit (compute hour) prices apply equally for both Lite and Pro for voice training. Check the pricing details here.	Per unit (compute hour) prices apply equally for both Lite and Pro. Check the pricing details here.

Responsible use of Custom Neural Voice

The access to Custom Neural Voice is limited in order to support Microsoft Responsible AI principles.As part of Microsoft's commitment to responsible AI, we are designing and releasing Custom Neural Voice with the intention of protecting the rights of individuals and society, fostering transparent human-computer interaction, and counteracting the proliferation of harmful deepfakes and misleading content. Registration with your use case is required for access to some features. Only customers managed by Microsoft, meaning those who are working directly with Microsoft account teams, are eligible for access.Learn how to apply for full access

How to create a professional Custom Neural Voice

1. Apply for accessLearn about responsible use of AI and apply for full access to CNV with your use caseApply for access

2. Design voiceDevelop a voice persona that defines the overall sound and emotional tone for your use caseDesign voice

3. Prepare scriptsPrepare scripts for your use case or utilize our sample scriptsPrepare scripts

4. Record voiceRecord samples and a voice talent statement at a professional recording studioRecord voice

5. Train voiceCreate a Pro project, upload recordings and scripts, train, test then deploy the voice

6. IntegrateUse the voice in your apps with Speech SDK, or create contents with the Audio Content Creation toolView quickstart

Sign in to try a Custom Neural Voice LiteTraining a Custom Neural voice Lite model requires Azure account and Speech resource (Standard S0).

Don’t have an Azure account yet? Sign up and get a free $200 credit, or learn more about creating an Azure account

Next steps

1. Sign in to try a Custom Neural Voice LiteSign in with an Azure account and create a Speech Services Standard (S0) resource to use Custom Neural Voice (CNV).

Don’t have an Azure account yet? Sign up and get a free $200 credit,or learn more about creating an Azure account

2. Apply to gain full access to Custom Neural VoiceCustom Neural Voice access is limited based on eligibility and usage criteria in order to support our Responsible AI principles. Review responsible use of AI and apply for full access to deploy a Lite voice or create a voice with Pro. This may take up to 10 business days for review.