For all the enthusiasm—not to mention dollar signs—buzzing around the art of podcasting over the last few years, it still suffers from one major flaw: The actual content of podcast episodes, like that of radio streams and other non-music audio found online, remains walled off and locked away inside impenetrable waveforms, far from the reaches of search engines and the content personalization algorithms that power what we read, see, and hear every day. But like so many things in life, there’s a startup looking to change this.
Audioburst is a Tel Aviv-based company with a mission so simple it’s hard to believe it hasn’t already been accomplished by a tech giant: It ingests millions of hours of audio per week, transcribes it, breaks it into chunks, and indexes it so it can be searched. Earlier this week, Audioburst launched a web search tool that lets you scour bits of audio from thousands of sources like radio stations and podcasts. So not only can it search across many popular podcasts, but it’s capturing audio from radio broadcasts, which until now have been beamed through the sky to radio receivers and web streams, only to be forgotten. The results of Audioburst’s search are still a little rough, but the idea is super-promising.
“We’ve built a machine that constantly listens to a variety of audio that’s being broadcasted and uploaded as podcasts,” says Audioburst founder and CEO Amir Hirsh. “As it’s listening, we do things to the audio to make it easy to use.”
Devices like your laptop or smartphone are great at finding text-based information and even images online, but come up empty-handed when the answers we seek happen to be buried inside the ones and zeros of audio files.
It’s kind of insane, really: We’ve been able to quickly and easily search the vast expanse of the web for over 20 years, yet digital audio remains stuck in 1995. Imagine, though, if you could search for a news topic, a sports team, or anything else people say that gets recorded, and get results not as a list of blue links, but as a series of play buttons. And when you hit play, you’re taken directly into the snippet of audio that pertains to whatever it is you were looking for. That’s what Audioburst’s new search engine does. The results are not perfect or comprehensive, but it can be quite useful for finding mentions of specific people, places, and topics.
Say you’re interested in meditation, for example. A search for that brings up a recent episode of NPR’s “Fresh Air” featuring Robert Wright talking about his new book on Buddhism and mindfulness. You can click and listen to the first relevant snippet or zoom out and play the entire episode. Then there are a few other public radio shows discussing meditation, all of which seem timely and relevant. Very cool. Further down the page, however, is a news clip about a meditation instructor who was inexplicably killed by a police officer. Probably not what you were going for! In many cases, the search functionality is somewhat crude and it sometimes misses the mark, but it’s nothing a little time and machine learning can’t fix.
Its new browser-based search engine is just the latest interface for Audioburst’s technology. The company also offers a stand-alone audio transcription service, as well as an API that lets app developers build Audioburst’s audio library, search functionality, and personalization into their apps and voice-controlled devices. The company is also exploring potential integrations with connected cars.
The service’s integration into voice control platforms like Alexa, Siri, and Google Assistant seems like Audioburst’s most ripe and promising prospect. Imagine, for instance, if instead of Alexa’s robotic voice updating you on the weather—or pulling in a one-size-fits-all NPR news radio stream, Amazon’s Echo devices could weave in personalized, relevant audio snippets from actual broadcasts. Audioburst’s existing integrations with Alexa and Google Assistant aren’t quite that tight, but the potential here is huge, especially in the escalating voice platform war.