SpaceXAI Releases Grok Voice Transcribe 2.0 with 50% Lower Error Rate
SpaceXAI released Grok Voice Transcribe 2.0, a speech-to-text model that cuts the word error rate by nearly half compared to its predecessor while maintaining the same pricing. The model achieved the top accuracy ranking among 32 streaming speech recognition models on the Artificial Analysis benchmark. By delivering a major leap in transcription quality at no added cost, SpaceXAI lowers the barrier for deploying high-accuracy voice AI across customer support, media transcription, and automotive platforms. This price-performance improvement pressures rival speech AI vendors and accelerates the adoption of real-time voice interfaces. Batch transcription pricing remains unchanged at $0.10 per hour of audio, while real-time streaming transcription costs $0.20 per hour. Premium features such as speaker diarization, precise timestamps, and custom keywords are included without extra fees.
## BACKGROUND
Automatic Speech Recognition (ASR) technology converts spoken audio into written text, typically evaluated using Word Error Rate (WER) to measure accuracy. Unlike batch processing of pre-recorded files, streaming speech recognition processes audio in real time, enabling immediate text output for interactive applications like vehicle assistants and live support.