立即打开
“音乐大脑”的读心术

“音乐大脑”的读心术

Rob Walker 2012年10月22日
大家或许从来没有听说过Echo Nest,但是这家公司的技术却支持着目前最炙手可热的音乐公司,比如Spotify、Vevo和MTV等。它建立的音乐数据库就好比一个能够理解音乐的大脑,成为音乐播放列表、相似歌手推荐等基本功能的灵魂。

    一切都是从一次争论开始的。两个博士生,特里斯坦•耶汉和布莱恩•惠特曼,在麻省理工学院(MIT)的媒体实验室(Media Lab)里相遇了。两人都是业余音乐人,而且他们都对一项技术非常感兴趣,也就是如何根据听众的品味向他推荐音乐。而且二人都相信,目前流行的“协作筛选”技术离实现这个目标差得还非常远。那么他们的分歧在哪在呢?耶汉的研究方向主要关注如何教会电脑捕捉音乐的声音元素,而惠特曼的研究方向主要关注音乐的文化和社会元素。他们将各自的研究方法结合在一起后,Echo Nest公司也就应运而生了。Echo Nest是目前最重要的数字音乐公司之一,只是很少有人听说过它。

    Echo Nest从2005年开始着手建立一个庞大的数据库,它就像一个能理解音乐的大脑,如果你喜欢当红嘻哈音乐巨星坎耶•维斯特,它就会向你推荐另一个嘻哈歌手德雷克的音乐。这听起来似乎跟音乐网站Pandora的服务很像。的确如此,不过Echo Nest的规模更大,而且它的服务并不直接面向听众,而是把数据卖给实力雄厚的流行音乐服务商和广播电台。

    Echo Nest的总部位于马萨诸塞州的萨默维尔市,这家公司融合了两种顶尖技术:首先,电脑程序可以分析歌曲的音调和节奏等基本元素。(Echo Nest的客户向公司提供进入歌曲文件的接口)与此同时,还有一个像搜索引擎一样的机制会在网络上收集关于该乐曲的微博和评论,比如“这首歌适合跳舞”、“这个歌手跟鲍伯•迪伦有一拼”等。两种技术结合在一起,赋予了每首歌曲一个独特的电子指纹,同时总结了歌曲的音乐和文化属性。通过这种方法,Echo Nest已经收集了200多万名歌手的3,400多万首歌曲的数据。Echo Nest公司内部的人管这个日益庞大的数据库叫“知识库”。

    Echo Nest的首席执行官吉姆•卢切斯于2007年加盟这家公司,帮助他们把这个创意推销给广大开发者和企业。一般来说,客户会定期向该公司支付许可费用,同时还会根据用户的人数,再支付一笔浮动的费用。早期阶段,采用Echo Nest技术的一个重要客户是音乐公司Spotify。2009年,正当这家流媒体服务公司风靡欧洲的时候,Spotify选择了Echo Nest的技术用于自己的播放列表功能,之后各种订单纷至沓来。比如Clear Channel公司就把这项服务用于它的iHeartRadio音频播放器;MTV也把这项服务用于它的Music Meter应用,这个应用每天都会更新最热歌手的排行榜。Echo Nest的客户中既有诺基亚(Nokia)、英特尔(Intel)这样的大牌企业,也有Raditaz和Rdio这样名不见经传的小型创业公司。音乐视频广站VEVO的总经理弗莱德•桑塔匹亚称:“他们是这个领域里做得最好的。”VEVO公司今年年初也通过Echo Nest的技术,添加了一个“相似歌手”功能。

    Echo Nest获得欢迎的原因,不仅仅是因为其数据库的规模。像谷歌(Google)和Facebook一样,这家公司也不遗余力地讨好开发者。Echo Nest还提供了一个设计得非常简捷的应用程序界面(API),让其它程序也能通过这个API调取Echo Nest的数据。

    在线流媒体服务公司MOG的创始人兼首席执行官大卫•海曼与Echo Nest开始合作是在2009年,当时,他的公司推出了一个高级订阅功能。除了Echo Nest的“知识库”之外,更令海曼印象深刻的,Echo Nest的服务可以非常轻松地与他自己的数据进行映射。换句话说,由于其他工程师能够将Echo Nest的技术与自己的工作融合,Echo Nest将潜在的竞争对手转化为了付费客户。

    此外,Echo Nest还推出了音乐应用“hack days”来提高该技术的开放性,并且还允许开发者免费使用该技术进行非营利性试验。位于澳大利亚的编程人员大卫•麦金尼指出,这项战略使Echo Nest成了那些想“打造新的音乐体验”的企业家的航空母舰。正是通过那些借助Echo Nest技术进行的试验,麦金尼拉来了投资者,成功建立了一家名叫Discovr的创业公司。

    It began with an argument. Tristan Jehan and Brian Whitman met as Ph.D. candidates at MIT's Media Lab. Both were amateur musicians passionate about the ways technology might recommend songs based on a listener's tastes. Both were convinced that "collaborative filtering," a trendy means of achieving that goal, was woefully inadequate. Their disagreement? Jehan's research focused on teaching computers to capture the sonic elements of music, while Whitman's studied the cultural and social components. In combining the two approaches they created the Echo Nest, one of the most important digital music companies few have heard about.

    Starting in 2005, they set about creating a vast database, a music brain that, based on your interest in Kanye West, can suggest you check out rapper Drake. Sound like Pandora? It's similar -- but on a massive scale. And instead of making its service directly available to listeners, the Echo Nest markets its data to power popular music services' custom playlists and radio stations.

    The Somerville, Mass., startup combines two types of cutting-edge technology: A computer program analyzes songs for their fundamental elements such as key and tempo. (The company's customers provide access to song files.) Meanwhile, a searchlike system crawls the web collecting what people are tweeting and commenting about music -- "that song is danceable," "that singer is Dylanesque." The two sets of data are combined to give every track a digital fingerprint containing its musical and cultural attributes. The Echo Nest has gathered data in this way for 34 million tracks by more than 2 million artists. Internally, the ever-growing database is referred to simply as "the knowledge."

    CEO Jim Lucchese joined the company in 2007 to help market the idea to developers and businesses. Clients would pay the company a recurring licensing fee, plus a fluctuating fee based on the number of users. An important early customer was Spotify. In 2009, when the streaming-music service was storming Europe, Spotify licensed Echo Nest technology for its playlist-building function. A series of deals followed, including one with Clear Channel for its online radio player iHeartRadio and another with MTV for its Music Meter app, which generates a daily ranking of hot new artists. Along with giants Nokia (NOK) and Intel (INTC), smaller music startups such as Raditaz and Rdio signed up in droves. "They're the best there is at what they do," says Fred Santarpia, general manager of music-video site VEVO, which added an Echo Nest-powered "similar artist" option earlier this year.

    It's not just the size of the database, of course. Much like Google (GOOG) and Facebook (FB), the venture has become successful by relentlessly courting developers. It provides an API -- a sort of front door that makes its data usable by other programmers -- explicitly designed to be simple.

    David Hyman, founder and CEO of online streaming service MOG, started working with Echo Nest in 2009 when his company launched a premium subscription option. Besides "the knowledge" itself, what impressed him was how easily it could be mapped onto his own data. In other words, Echo Nest's ability to meld with the work of other engineers has transformed would-be competitors into paying customers.

    The company promotes this openness by hosting music-app "hack days" and giving developers free access to its technology for noncommercial experimentation. The strategy is making it the "mothership" for entrepreneurs looking to "create new musical experiences," explains David McKinney, a coder in Australia whose experiments led to the creation of an investor-backed startup called Discovr.

  • 热读文章
  • 热门视频
活动
扫码打开财富Plus App