Soundhound, known for its popular music recognition service, will soon launch a search app driven by voice recognition technology, pitting the Californian company against major players such as Google.
Dubbed ‘Hound’, the new app for iPhone and Android devices allows users to speak search terms to deliver tailored results in a specific category. Initially, Hound will only allow users to search for information on music and bands, but the company plans to broaden search into other categories.
According to sales and marketing vice president Katie McMahon, voice-controlled applications are set to become the next major trend in mobile technology.
“The next revolution to happen after the touch and tap phenomenon is going to be interacting with the device not with our fingertips, but with the air out of our lungs,” McMahon told Macworld Australia. “You’re just going to say it and – whoosh! – it all comes up!”
Other companies seem to agree. Google has been developing voice search for some time, recently introducing the technology to its translation services, and Apple is widely understood to be working closely with Nuance on an undisclosed voice product.
Introducing a discrete, category-based, walled-garden approach to search puts Soundhound in direct competition with the search giant Google. McMahon conceded that drawing such an inference is a ‘fair parallel’.
She said the time is right for the company to “nail this brand new user habit”.
“Just as in 2002 the user had no concept of how delightful a touchscreen would be, right now Soundhound Inc looks at the marketplace and (knows) the timing is right because it’s virgin territory.”
Voice recognition has been an elusive goal for developers since the days of early computing. In the past, the technology has been hamstrung by roundly terrible comprehension.
“Voice recognition as a concept and a technology has been around for 30 years, but it’s just such rubbish it’s not delightful,” conceded McMahon. “We all cringe when we’re stuck on the phone with an automated voice-recognition banking service or your airline. It’s just a disaster. I end up shouting ‘OPERATOR OPERATOR OPERATOR’ out of frustration.”
Previous voice recognition technology worked by analysing audio, converting words into text and checking those against a database. This method, apart from being far from perfect at comprehension, propagated errors down the line. Soundhound, however, believes it’s found a solution.
“In lab test scenarios, it’s 99.9 percent accurate,” McMahon claimed.
Building on its experience in the music recognition field – and its large database – engineers at Soundhound developed a ‘sound-to-sound’ method that matches the voice against a catalogue of similar audio. Once it understands the request, Hound matches the search terms against its own set of information and delivers previews of the song, cover art, album details, lyrics, videos, tour dates and an option to buy the song instantly.
While initially limited to music, Soundhound believes the technology can be applied to other categories of information.
“We’ll roll it out in music, but our roadmap and our intent is future versions will have enhanced categories that will relate to daily things you would hit your mobile phone for,” said McMahon. “A company like this could not have such ambitions realistically in 2005. But in 2011 we absolutely can have a vision and move forward in revolutionising how people interact with their connected devices.”
Hound launches in the United States this week and a local release is expected in the near future, with the company currently working on implementing voice recognition for non-US accents.
“We’ve actually tested it quite extensively on some Australian accents and it works really well,” McMahon said. “But we’re going to take our time with the roll-out. I’d be concerned that somebody in Glasgow might have a super-thick accent…”