科大讯飞在今日下午的发布会上宣布了一项重大的技术突破，其星火语音大模型在首批37个主流语种上已整体超越OpenAI公司推出的Whisper-large-v3。这标志着我国在智能语音技术领域继续保持国际领先水平。
星火语音大模型由多语种语音合成，其最大的亮点在于对多语种的支持。在首批40个语种中，星火语音大模型的平均MOS分绝对提升了0.25，MOS分达到了4.5，拟人度也高达83%以上。这些数据显示，星火语音大模型不仅在技术上实现了突破，而且在用户体验上也做到了极致。
作为我国智能语音技术的领军企业，科大讯飞一直致力于推动我国智能语音技术的发展。星火语音大模型的发布，不仅是科大讯飞技术实力的体现，也是我国智能语音技术发展的一个重要里程碑。
With the release of the Spark Speech Grand Model, iFLYTEK has achieved a significant breakthrough in the field of intelligent speech technology, surpassing OpenAI’s Whisper-large-v3 in the first batch of 37 mainstream languages. This marks China’s continued leading position in the international market for intelligent speech technology.
The Spark Speech Grand Model, capable of multilingual speech synthesis, is the highlight of the launch. It has achieved an average absolute improvement of 0.25 in the MOS score across the first batch of 40 languages, reaching a MOS score of 4.5 and a human-like degree of over 83%. These figures demonstrate that the Spark Speech Grand Model not only represents a technological breakthrough but also delivers an exceptional user experience.
As a leading company in China’s intelligent speech technology sector, iFLYTEK has been committed to promoting the development of this field. The release of the Spark Speech Grand Model is not only a demonstration of iFLYTEK’s technical strength but also an important milestone in the development of China’s intelligent speech technology.
【来源】https://news.mydrivers.com/1/961/961266.htm
Views: 1