Accessible Speech Recognition Tech Startup Voiceitt Raises $4.7M
Atypical speech recognition technology startup Voiceitt has closed a $r.7 million funding round led by AMIT Technion along with Cisco Investments and Third Culture Capital. Voiceitt’s software enables voice assistants and related tools to understand people with non-standard or impaired speech through a mobile app and API integration. Unexpected enthusiasm led to an oversubscription of the funding round, which the company has now upped to a $10 million goal.
Voiceitt’s technology works to train speech recognition models that can understand people whose speech differs significantly from that used to train the most popular voice assistants. The iOS mobile app and API work with the user to learn their specific speech pattern, pronunciation, and subtle elements like non-verbal sounds. which may have been affected by illness or injury. Ongoing use improves the AI’s comprehension though it can produce a rough working model within a few minutes of repetition. The personalized voice AI can then repeat what the user said, interpreting their words for other people or even other voice assistants.
Voiceitt is particularly well suited for Alexa devices as the startup worked with Amazon’s Lab126 to deeply integrate it into Alexa to the point it’s not a separate skill, just built into the voice assistant to silently interpret non-standard speech so that Alexa can understand it without an out-loud repetition. Voiceitt was part of the 2018 Amazon Alexa Accelerator and eventually became an Amazon Alexa Fund portfolio company.
“Our next generation technology, by expanding our current product offering and supporting spontaneous speech, enables individuals with speech and motor disabilities to perform everyday tasks independently – whether it is composing texts, making social media posts, or communicating with smart home devices,” Voiceitt CEO Danny Weissberg said. “The great speed at which we continue to accelerate the algorithm development, vastly expand our large voice recordings database, and ultimately scale successfully is thanks to support from strategic investors like AMIT, Third Culture Capital and Cisco Investments, who share our vision to make voice AI accessible to people with non-standard speech.”
The decade-old startup was born out of has raised $20 million in investment, including a $10 million Series A round two years ago led by Viking Maccabee Ventures, whose CEO, Karl Anderson, is also Voiceitt’s chairman of the board. Most recently, Voiceitt released a beta test of an upgraded AI capable of understanding and translating people with speech disabilities as they speak.
Interest in making voice technology more accessible has grown among the major voice AI developers of late. Amazon, Apple, Google, Meta, and Microsoft joined with the University of Illinois (UIUC) a couple of months ago to form the Speech Accessibility Project with that goal in mind. The project is collecting and analyzing speech samples from people with a broad array of speech impairments specifically to train new AI models that can understand atypical speech.
The Speech Accessibility Project grew from previous work by tech companies to widen speech recognition to encompass those with disabilities. Google began testing Project Relate for that purpose, using data from Google’s Project Euphonia, while Apple has been researching techniques to help Siri know when someone is stuttering to compensate and ensure it doesn’t interrupt or misunderstand the user.