科技资讯

分解后再合成,浙江大学联手字节跳动推出语音合成系统Mega-TTS

发布日期:2023-06-26    点击次数:127
文章来源于品玩GenAI,作者大模型机动组浙江大学研究团队近日联手字节跳动,推出了全新 Zero-shot语音合成系统Mega-TTS。当前的语音合成系统通常是通过自回归语言模型或扩散模型来生成语音,这会忽略语音的内在本质,导致输出结果可能出现劣质或不可控的情况。该研究团队通过将语音分解为内容、音色、韵律等不同的属性,并针对每个属性进行建模,为此他们设计出了全新的Zero-shot语音合成系统Mega-TTS。通过使用大规模的野生数据进行训练,并以不同的方式来对不同的属性进行建模。

","gnid":"952535465b745f52a","img_data":[{"flag":2,"img":[{"desc":"","height":"720","title":"","url":"http://p1.img.360kuai.com/t01390ba19c14e58341.jpg","width":"1080"},{"desc":"","height":"1920","title":"","url":"http://p0.img.360kuai.com/t0156959cf05830e769.jpg","width":"1080"},{"desc":"","height":"748","title":"","url":"http://p2.img.360kuai.com/t016331fc30f147ee96.jpg","width":"1080"}]}],"original":0,"pat":"art_src_1,fts0,sts0","powerby":"pika","pub_time":1686191710000,"pure":"","rawurl":"http://zm.news.so.com/b7c2b3b58f1140cfff54b20cd09b9207","redirect":0,"rptid":"235ea931c6db6861","rss_ext":[],"s":"t","src":"品玩","tag":[{"clk":"ktechnology_1:mega","k":"mega","u":""},{"clk":"ktechnology_1:浙江大学","k":"浙江大学","u":""},{"clk":"ktechnology_1:字节跳动","k":"字节跳动","u":""}],"title":"分解后再合成,浙江大学联手字节跳动推出语音合成系统Mega-TTS","type":"zmt","wapurl":"http://zm.news.so.com/b7c2b3b58f1140cfff54b20cd09b9207","ytag":"科技:人工智能:AI技术","zmt":{"brand":{},"cert":"优质科技领域创作者","desc":"有品好玩的科技,一切与你有关。","fans_num":9264,"id":"2991151609","is_brand":"0","name":"品玩","new_verify":"7","pic":"http://p5.img.360kuai.com/t019112a1b3e04850a2.jpg","real":1,"textimg":"http://p9.img.360kuai.com/bl/0_3/t017c4d51e87f46986f.png","verify":"0"},"zmt_status":0}","errmsg":"","errno":0}

上一篇:36氪首发|「图灵机器人」完成近亿元A轮融资,专注工业机器人研发与应用
下一篇:DeepMind人工智能创造出比人类快70%的排序算法