Upload
realspeaker-lab
View
102
Download
0
Embed Size (px)
DESCRIPTION
Citation preview
RealSpeaker - audio-visual enhancement to - RealSpeaker audio-visual enhancement to speech recognition system
PROBLEM
Inability to Suppress AmbientNoise (audio is not reliablesource of information)
High Cost of Voice RecognitionApplications (Nuance licensesaverage costs from $100 to$1000)
Issues with Accuracy (withaccuracy 60-70% their paintfullto use)
Low Level of Security inSpeaker Verification
Users must speak in unnaturalfashion using fragmentedspeech (the problem withusability)
EXISTING ALTERNATIVES
Keyboard typing
DragonDictation for PC or Mac(very expensive costs)
Google Speech recognition(free using by default onAndroid OS - only for short voicecommands)
Windows speech recognition(free using by default onWindows OS - the problem withaccuracy - only for short voicecommands)
Siri voice assistent on iOS (freeusing by default on Iphones -only for short voice commands)
SOLUTION
RealSpeaker uses additionalvideo information, which allowsto improve voice recognitionaccuracy by at least 20-30 percent.
More safety becauseRealSpeaker have function ofaudio-video verificationspeaker's speech from theoverall speech flow
RealSpeaker cheaper thanNuance. Licences costs from$25 to $90
RealSpeaker have functions ofvoice editing and sending -more usability than Nuance
UNIQUE VALUEPROPOSITION
The average typing speed is 33words a minute. By the means ofRealSpeaker can record over100 words a minute
We have paid users, workingprototype and good traction
Multilanguage - (over 13languages supported today)
Enter text of any length with thevoice and video withoutkeyboard at any text editor orwebsite (Notes, Facebook,Skype, Evernote, E-mail,Microsoft Office etc.)
HIGH-LEVEL CONCEPT
RealSpeaker - audio-visualenhancement to speechrecognition systems
OpenCV - video processinglibrary originally developed byIntel
Nuance - speech recognitioncompany
Google Voice Search - speechrecognition engine
CMU Sphinx - speechrecognition system with OpenSource code
GoogleGlass SDK - videoprocessing library
Kinect SDK - video processinglibrary
UNFAIR ADVANTAGE
US Patent 13/942,689: “Systemof video enhancement for audiospeech recognition solutions toimprove the accuracy of audiospeech recognition due to theanalysis of speaker lipmovements”
Our team is supported by suchorganizations and institutes asMicrosoft Seed Fund, Skolkovo,Startobaza, Kazan IT-Park. Wehave NDA with Samsung, LG,Toyota, Itouchu.
We exists almost 2 years andhave a team of 10 people - ourworking place is based in Kazan(Russia) - cheap place withgood professional
Our technology can be integrateat any electronic devices
We have own database - videohow its work -http://youtu.be/TQaVWTqGCjs
CUSTOMER SEGMENTS
Adults group: disablity people
Professional segment: SEO,journalists, writers, bloggers,students, teachers, coachers,mentors, research specialists,focus groups
Active segment: businessmen,teenagers, geeks
According to TechNaviocurrently, only 15-20% of thespeech recognition marketpotential is used - need only tocreate customer product withhigh accuracy speechrecognition
EARLY ADOPTERS
Bloggers, journalists, roboticgeeks, journalists - our firsttesters
We have released beta versionof our product for Windows OS,which is currently in use byabout 50k users, out of which 2kusers are paid users
KEY METRICS
Funding: - Current Burn Rate -$1M - Seeds Round: $0,5M in2012-2013 o $0,3M – in 2012 o$0,2M – in 2013 - Seeking 1stRound: $1,0M
about 50k users, out of which 2kusers are paid users (December2013)
CHANNELS
Viral channelwww.realspeaker.net (prizes onour site)
Social servises: YouTube,Facebook or others
Free Torrents (spread trialversion of RealSpeaker)
Software Vendors (MailRuGroup , Digital River)
Lean Canvas is adapted from The Business Model Canvas (BusinessModelGeneration.com) and is licensed under the Creative Commons Attribution-Share Alike 3.0 Un-ported License.
COST STRUCTURE
Cost per license in time:
$25 for 3 months
$30 for 6 months
$37 for 1 year
$90 unlimited version
Integration at any service - royalty from sales
REVENUE STREAMS
Revenue: - 2012: $0,025M - 2013: $0,1M - 2014: $0,5M - 2015: $15M - 2016: $100M
B2C segment
Business Model: Try & Buy Free version – can recognize speech to text for 3 days; 4 %conversion. 50 k free users, 2 k paid users in December of 2013
B2B
We have NDA with Samsung, LG, Toyota, Itouchu.