In a video interview with Geektime at Microsoft’s Think Next conference in Tel Aviv, the company said it wants to let you find and share your favorite snippets, finally bringing the power of social media to radio
You’re driving in your car, when you hear a hilarious joke on the radio that brings tears to your eyes. You reach your destination and try to search for the joke online, but to no avail. You’d love nothing more than to share the audio clip with your friends, but there’s no way to do it.
A pre-launch Israeli company called AudioBurst wants to change that.
“The radio industry today produces a lot of premium content,” the company’s CEO, Amir Hirsh, told Geektime’s Laura Rosbrow in a video interview at Microsoft’s annual Think Next conference in Tel Aviv.
“They invest a lot of money and effort into it. There’s only one problem. They’re broadcasting it live and if you didn’t hear it right then, it vanishes into thin air.”
AudioBurst plans to record radio content, store it permanently and transcribe it. The text of radio shows will then be made available to search engines.
All radio shows?
Yes, Hirsh told Geektime in a post-event interview. AudioBurst will start with the U.S., recording all the content on over 6,000 talk radio channels, but eventually, “We will cover all radio, in all languages, all around the world, which will be searchable by topics, categories, mood, location and context.”
Unlike platforms such as SoundCloud, TuneInRadio and Stitcher, AudioBurst will also slice and dice audio content into snippets or “bursts” so you can share that one-minute joke with your friends on Facebook or WhatsApp.
“You can listen to an interview, a traffic report or weather report. In our ADD kind of world, you don’t need to listen to the whole hour. You can listen to that specific burst that was interesting to you.”
In the world of media buzz, audio sometimes seems like the ignored middle child sandwiched between text and video. But in fact, audio is the perfect medium to accompany mundane tasks that require some, but not all of our attention: driving, exercising, cooking, walking, even ironing our clothes. As broadband connections proliferate, Internet radio audiences are on the rise. In the United States, the number of unique monthly podcast listeners has reached 75 million, up from 25 million five years ago.
“Technologies have come and gone, but radio kept its popularity,” said Hirsh. “Over 90% of Americans listen to the radio at least once a week.”
In fact, Hirsh said the idea for AudioBurst came about when he couldn’t find something he was looking for on the radio.
“Our company has been offering online video transcription services for Fortune 500 companies for years, so when one day we were looking for a specific radio interview of a friend of ours, it seemed strange we couldn’t find it anywhere. The words spoken were not saved anywhere.”
How will they make money?
“We have two income models. The first is in revenue sharing with the applications and services developed on top of our online repository and APIs, much like Facebook and other similar ecosystems. The second is by generating revenues from apps that we will develop ourselves, mainly aiming at ad-supported consumer listening,” Hirsh explained in Geektime’s video interview.
AudioBurst’s online repository and developer APIs will not be ready until the end of 2015. But in a few months they plan to launch a website “that will allow users to see everything that is hot and happening on the radio.”
In addition, AudioBurst will launch a Radio Alert app later this year that will notify users any time a topic of interest to them has been mentioned on the radio.
So if you are obsessed with, say, horticulture and want to listen to back-to-back horticulture talk shows while driving home from work, this app will finally make that possible.
The holy grail of speech recognition
When was the last time you tried to use speech transcription software? Suffice it to say that as a journalist, I still transcribe most of my interviews manually. Even the best software out there has to be trained to interpret your voice, and still makes a lot of mistakes. Has AudioBurst overcome this major technological hurdle?
Hirsh claims that Audio Burst is in the listening business, not the transcription business, so its transcriptions don’t have to be perfect, just good enough to search.
“Our trainable speech-to-text engine improves over time and is constantly supplemented with extra NLP layers of topical extraction. It provides a premium level of accuracy that is optimized for our platform to search upon.”
But users don’t see any of this. Once they search and find the desired clip, they can sit back and listen and enjoy.