SpaceX AI推出了语音识别模型Grok Voice Transcribe 2.0，相比前代版本实现了准确度的显著提升1。新版本的准确度达到1.0版本的两倍，但定价保持一致，批量转录保持在0.10美元/小时，流式转录为0.20美元/小时1。
该模型在第三方评测中表现突出，在Artificial Analysis公开排行榜的32个流式模型中排名第一1。技术上，Grok Voice Transcribe 2.0支持19种语言的自动检测与中途语言切换1，多语言短语识别的错误率从20.6%大幅下降至6.8%1。目前已有Atlassian Loom等客户采用该模型1。SpaceX AI宣布，旧版本Grok Voice Transcribe 1.0将在未来数周内停用1。
SpaceX AI has released Grok Voice Transcribe 2.0, a significantly upgraded version of its speech recognition model that achieves twice the accuracy of its predecessor while maintaining the same pricing structure 1. The new model is priced at $0.10 per hour for batch transcription and $0.20 per hour for streaming transcription 1. In performance benchmarks, Grok Voice Transcribe 2.0 ranks first among 32 streaming models on the Artificial Analysis public leaderboard 1.
The enhanced model supports automatic detection and mid-stream switching across 19 languages, with error rates in multilingual phrase recognition dropping from 20.6 percent to 6.8 percent 1. The technology has already been adopted by customers including Atlassian Loom 1. SpaceX AI has announced that Grok Voice Transcribe 1.0 will be deprecated in the coming weeks 1.
评论
还没有评论，欢迎留下第一条。