While there are already speech-to-text apps that allow deaf users to read what other people are saying, they're often thwarted by background noise or multiple people speaking at once. SpeakSee, however, is a new multi-mic system that reportedly works much better.
SpeakSee was created by Dutch entrepreneur Jari Hazelebach (working with company co-founder Marcel van der Ven), who grew up with two deaf parents who often had difficulty following family conversations.
The basic setup consists of three Wi-Fi wireless mics (although up to nine can be accommodated), a mic docking station/charger, and an iOS/Android app. Each mic is clipped to the shirt of a person involved in a conversation, and utilizes a "beam-forming" directional system to pick up their speech while not picking up background sound, or the speech of other people.
The mics transmit the audio to the docking station, which can be located up to 20 meters away (66 ft). It processes the speech picked up by the different mics, then relays the audio data via Wi-Fi to the app on the deaf user's smartphone. That app transcribes the data into text that is displayed onscreen in real time, with different colors and names indicating which person is speaking.
Transcripts of conversations can be saved on the app, and then subsequently searched using keywords. Using an integrated 3.5-mm jack, it's also possible to plug devices such as smartphones or TVs directly into the docking station, allowing their speech audio to be analyzed and transcribed.
Over 120 languages and variants are supported.
Battery life for the mics ranges from two to four hours per charge, depending on how much talking the wearer is doing. The docking station can also be used while not plugged into an outlet, with its battery good for about eight hours of use.
SpeakSee is currently the subject of an Indiegogo campaign, with a pledge of US$249 getting backers a 1-mic solo kit, and $379 getting them a 3-mic basic kit, if all goes according to plans. The estimated retail price for the basic kit is $699.
It's demonstrated in the following video.
