Tuning resources

You may further improve your speech output with custom lexicons and recorded audio prompts with help from Microsoft utilities.

Custom lexicons

Microsoft provides instructions for creating a lexicon to improve pronunciation using the standard pronunciation lexicon specification. Full details are in the Microsoft documentation: Custom lexicon  .

See also SSML elements: Lexicon in this documentation for a simple XML file that provides custom pronunciations using the sample client.

Microsoft also offers a validation tool available from GitHub: Cognitive-Speech-TTS: CustomLexiconValidation  .

Audio prompts

Your SSML synthesis requests can include audio prompts using the <audio> element.

You may create audio prompts using Microsoft’s Audio Content Creation tool. These recordings can then be exported to your audio library for a seamless storage, management, and integration with your clients using Azure blob storage. Alternatively, you may download these prompts and upload them to third-party storage.

For full instructions, see the Microsoft documentation: Speech synthesis with the Audio Content Creation tool  .

Voice testing demo page

Microsoft offers a demo page that lets you experiment with different speech attributes such as voice style, rate, pitch, volume, pronunciation, and breaks. See Audio Content Creation  .