AI 导读
在首次部分转录上，Grok Voice Transcribe 2.0 在语音结束后 0.49s 达到 3.4% WER，在准确率上略领先于 Muse Voice Transcribe 和 ElevenLabs Scribe v2 Realtime，后两者均为 3.6%，但在速度上有所取舍。Grok 也比使用外部端点的 Cartesia Ink-2（4.0% WER，0.07s）以及 AssemblyAI U3.5 Realtime Pro 更准确，后者的 Max Accuracy 和 Min Latency 模式均达到 4.0% WER。
正文
On First Partial Transcript, Grok Voice Transcribe 2.0 achieves a 3.4% WER at 0.49s after end of speech, just ahead of Muse Voice Transcribe and ElevenLabs Scribe v2 Realtime on accuracy, both at 3.6%, while trading some speed. Grok is also more accurate than Cartesia Ink-2 with external endpoints at 4.0% WER and 0.07s, and AssemblyAI U3.5 Realtime Pro, whose Max Accuracy and Min Latency modes both achieve 4.0% WER.
来源：@ArtificialAnlys · x.com