SpaceXAI Launches Grok Voice Transcribe 2.0 Speech-to-Text Model

SpaceXAI released Grok Voice Transcribe 2.0, a speech-to-text model optimized for high accuracy in noisy, real-world environments. Built on the audio foundation used for Grok Voice, the model is integrated into Tesla vehicles and various customer-support systems.

The model currently ranks first for accuracy among 32 streaming models on the Artificial Analysis leaderboard. Version 2.0 introduces improvements over its predecessor in multilingual accuracy, the handling of short phrases, and performance on telephony audio. Technical capabilities include speaker diarization, key term biasing, and support for both batch and streaming transcription.

Atlassian Loom reported that the new model outperforms its previous transcription solution. SpaceXAI maintained existing pricing for the service, charging $0.10 per hour for batch transcription and $0.20 per hour for streaming.