More than nine years of top-secret work have gone into creating this product demo. And it’s one of the better ones I’ve seen.
“Show me hotels in Seattle for Friday, staying one night,” SoundHound CEO Keyvan Mohajer orders the Hound app on his Android phone. I was leaving for Seattle the day after our meeting, so Mohajer is using my trip as an impromptu scenario to show me how his voice search is smarter than Siri and Google Now. In a few seconds, 10 results appear on a map.
“Show only the ones costing less than $300,” Mohajer refines, explaining that the slight delay of three to four seconds — not much, considering it’s travel search — is because this feature is powered by a relationship with Expedia.
Then he starts listing more parameters, without slowing down: “Show only the ones that have three or four or five stars that are pet friendly, that have a gym and a pool, within 4.5 miles of the Space Needle.” The app gets that right away, and refines the results.
Then Mohajer ups the stakes by changing the original parameters. “What if I check in on Thursday, and stay for two nights?” Without the usual annoyance of a back button resetting the page, we see results for the same complicated search on different dates. Siri and Google Now definitely don’t do this (I tried; they both gave me unsorted lists of hotels in Seattle).
Mohajer explains the difference is that his technology is “speech to meaning” instead of “speech to text.” By combining speech recognition and natural language understanding into one engine — that was the big nine-year effort — Hound can be faster than the voice search competition and understand more compound queries.
And this is not just a hotel search app — it’s an app where you can press a button and ask just about anything: Weather, stocks, time zones, even a complex mortgage calculation or currency conversion.
To get the Hound app to market, Mohajer did not follow the typical entrepreneurial path. Startups, especially mobile app startups, don’t enjoy nine years of R&D before launch.
You may have heard of SoundHound before — it makes the music recognition app of the same name, which is much like Shazam except that it recognizes songs even when you’re humming them. It’s a neat magic trick. It’s been installed more than 260 million times.
Mohajer founded SoundHound in September 2005 with the goal of creating a way for humans to talk to computers, and computers to talk back. He raised some money from WaldenVC, TransLink Capital and Global Catalyst Partners. But the big ideas were a long way off from coming to reality. So while the main technology was fermenting, SoundHound set up an interim business — the music recognition app — which supports the company’s operations through ads and fees from music stores when people buy songs after searching for them.
The new Hound app won’t be broadly available yet — it’s being released invite-only on Android. (iOS is in the works.) It’s a steep uphill battle given that both Google and Apple have their own voice search offerings built into their operating systems.
That’s not the only challenge: Will Hound actually work as well as it demos? Even if people download the Hound app and remember to open it, will they expand their imaginations to ask questions outside of the Tarzan-style keyword searches we’ve all become familiar with? Will they take the time to learn about Hound’s 50 different specialized domains (Siri has something like 20)? That’s a tough job for a startup best known for its humming magic tricks.
Mohajer admits those challenges are real — but he reminds me how good the demo is. And he says the consumer app isn’t SoundHound’s only bet. “Our vision is: Everything could be Houndified.” That is, the Hound technology will be available to developers as a tool called Houndify, with a “non-prohibitive” price that includes a free tier. So maybe even if I don’t use Hound itself, the next time I use somebody else’s hotel search, it will work like Mohajer’s demo.
This article originally appeared on Recode.net.