xAI releases Grok Voice Transcribe 2.0 speech-to-text model

xAI, also referred to as SpaceXAI, has released Grok Voice Transcribe 2.0, a new speech-to-text model. The company claims the new version is twice as accurate as its predecessor, Grok Voice Transcribe 1.0, while maintaining the same pricing structure. The model is designed to handle difficult real-world audio conditions, such as noisy environments, competing voices, and local accents.

On the public Artificial Analysis leaderboard, the model ranks first for accuracy among 32 streaming models. The technology is built on the audio foundation model used for Grok Voice, which currently powers customer-support calls, video narration, and voice assistants in Tesla vehicles. The model features automatic language detection and can handle mid-recording language switches.

In terms of pricing, the model is set at $0.10 per hour for batch processing and $0.20 per hour for streaming. Features such as speaker labeling, word-level timestamps, and key term biasing are included at no additional cost. The model was trained on a dataset of live, multilingual audio recorded across diverse environments.