Introducing Real World VoiceEQ: Measuring the human quality of voice AI

A new benchmark called Real World VoiceEQ has been introduced to measure the human-like quality of voice AI interactions. It argues that existing benchmarks focusing on metrics like word error rate are insufficient, as they fail to capture nuances in tone, emotion, and context. The benchmark evaluates over 40 proprietary and open-source models across dimensions like Automatic Speech Recognition (ASR), Text-to-Speech (TTS), and Speech Understanding. Developed from over a million human ratings, it aims to provide a more holistic assessment of voice AI performance in real-world scenarios. Key findings suggest that models have become better at speaking than listening, and traditional benchmarks often overestimate their true capabilities.