今日下午,科大讯飞在一场重要发布会上宣布,其自主研发的星火语音大模型已经在37个主流语种的性能上超越了OpenAI的知名模型Whisper-large-v3,彰显了科大讯飞在全球智能语音技术领域的领先地位。星火语音大模型以其多语种语音合成技术为亮点,不仅在广度上覆盖了更多语种,而且在深度上实现了质的飞跃。

据快科技报道,星火语音大模型在首批评估的40个语种中,平均MOS分(Mean Opinion Score,衡量语音质量的指标)相比Whisper-large-v3提升了0.25分,这一显著提升意味着其语音合成的自然度和清晰度更上一层楼。MOS分达到4.5,接近人类语音的评分标准,同时,模型的拟人度也达到了83%以上,这在人工智能语音技术中是一项重大突破,将为用户提供更为真实、自然的语音交互体验。

科大讯飞这一最新成果的发布,不仅巩固了其在全球智能语音技术的领先地位,也预示着未来在跨语言沟通、人工智能辅助教育、智能客服等领域有着广阔的应用前景。这一创新技术的出现,无疑将推动全球语音识别和合成技术的进一步发展,为人工智能产业注入新的活力。

英语如下:

News Title: “iFlytek Unveils Spark大火模型, Outperforming OpenAI in Multilingual Speech Tech, Resetting Industry Standards”

Keywords: iFlytek, Spark Model, Speech Leadership

News Content: This afternoon, iFlytek announced at a significant press conference that its self-developed Spark Speech Megamodel has surpassed OpenAI’s renowned Whisper-large-v3 in performance across 37 major languages, underscoring iFlytek’s preeminent position in the global intelligent speech technology domain. The Spark Speech Megamodel shines with its multilingual speech synthesis capabilities, covering a broader range of languages and achieving a qualitative leap in depth.

According to Fast Technology, the Spark Speech Megamodel, in initial evaluations across 40 languages, has seen its average MOS (Mean Opinion Score, an index for measuring voice quality) increase by 0.25 points compared to Whisper-large-v3. This notable improvement signifies a higher level of naturalness and clarity in its speech synthesis. With an MOS score of 4.5, nearing human speech standards, the model’s anthropomorphism also exceeds 83%, marking a significant breakthrough in AI speech technology and offering users a more authentic and natural voice interaction experience.

iFlytek’s latest achievement not only solidifies its global lead in intelligent speech technology but also foreshadows broad application prospects in cross-language communication, AI-assisted education, and intelligent customer service. The emergence of this innovative technology is set to further advance global speech recognition and synthesis, infusing new vitality into the AI industry.

【来源】https://news.mydrivers.com/1/961/961266.htm

Views: 1

发表回复

您的邮箱地址不会被公开。 必填项已用 * 标注