xAI has released 21 new flagship voices for its Grok Voice service, expanding the available lineup beyond the original five options. These new voices are natively multilingual, supporting over 25 languages, and are accessible via the realtime Voice Agent API, the Text to Speech API, and the Grok Voice Agent Builder.
- The new voices are cast for specific roles including support, characters, commentary, advertising, and education.
- Users can control delivery using speech tags such as [pause], <whisper>, and <emphasis>.
- The original five voices (Ara, Eve, Leo, Rex, and Sal) have been retrained with an improved recipe to enhance naturalness in pacing, phrasing, and emphasis.
- Custom voice agents can be built using the Grok Voice Agent Builder, which also allows cloning a user's own voice from approximately one minute of audio.
The update aims to provide more versatile and natural-sounding options for developers building voice applications.