The latest speech model from SpaceXAI allows users to clone voices from two minutes of audio and provides text to speech in more than 30 voices across 25 languages. Available now on the fal platform, Grok Voice features a response time of 0.70 seconds and completes tool calls before the generated sentence ends.

Developers can use the model to build low latency agents designed to resolve customer service issues. Technical capabilities include transcription with word level timestamps, which helps automate the delivery of real time voice interactions for businesses using the hosted platform.