Last May, I had an opportunity to first meet and interview Klemen Simonic, Founder and CEO of Soniox. Back then, the Soniox team invented a novel self-learning AI algorithm that leverages vast amounts of available unlabeled audio and text to teach itself how to recognize complex speech patterns. This enabled Soniox to build an extremely accurate speech-to-text AI solution beating out market leaders like Amazon, Microsoft and even Google on speech recognition accuracy.
In a follow-up interview with Klemen this past week, I was thrilled to hear of their next breakthrough AI technology for voice and speech applications. Today, Soniox is releasing a new AI technology called Soniox Speaker AI.
Soniox Speaker AI is an artificial intelligence for recognizing speakers (“who said what”) based on audio information alone. Soniox Speaker AI has achieved an incredible 96% accuracy in recognizing speakers for all kinds of business applications. With an error rate of just 4%, Soniox Speaker AI is 5 times more accurate than Google, Amazon and Microsoft solutions, unlocking a new dimension for voice and conversational applications.
Watch Soniox Speaker AI Video here to learn more.
Knowing “who said what” is a fundamental element when building any kind of voice applications or conversational AI. Leading providers of speaker recognition, such as Google, Amazon and Microsoft incur on average an error of 25%, meaning a quarter of the speech is assigned to the wrong speaker, making recognized speaker information unusable for downstream application.
Attempting to develop conversational AI without knowing who asked a question or who gave an answer is infeasible and futile. Reliable speaker recognition is in fact a prerequisite for making any conversational AI application work. Examples include customer support, sales, recruiting and not mentioned medical conversations (who is the doctor and who is the patient). The speech analytics world has taken a giant leap forward with much needed and improved accuracy of Soniox Speaker AI.
As of today, Soniox announced two new products, Soniox Speaker Diarization AI, and Soniox Speaker Identification AI. Soniox Speaker Diarization AI transforms a single audio into a multi-speaker transcription. Soniox Speaker Identification AI identifies speakers based on voice profiles and assigns speaker names to sentences. To push technology envelope even further, both of these products support offline or batch processing as well as real-time and low-latency setting, making Soniox the first in the world to offer such capabilities.
To put things in context, the speech and voice recognition market is predicted to be worth over $31.82 Billion by 2025 according to Grand View Research, Inc.
Soniox offers world-leading speech recognition accuracy, and as of today, world-leading speaker recognition and identification AI. Any application using speech as input can improve their product by using Soniox Speech and Speaker AI. Soniox APIs can be used from virtually any programming language and platform.
How can you learn more about Soniox?
The company was founded in April 2020, in Redwood City, California and has headquarters in Foster City in the Bay Area and an engineering team in Europe.
Soniox offers speech AI services and products for enterprises, developers and consumers. Soniox has built both the Soniox web application and the Soniox mobile application (for iOS devices).
Why is this breakthrough important?
This means companies building voice recognition solutions can use the Soniox product offering and build more trusted and accurate solutions for their customers, and accelerate the quality of speaker identification.
I suspect there will be many offers flowing with this announcement, as Amazon, Microsoft, and Google like to swallow up innovations pretty rapidly. At the same time, I think Klemen and his leadership team have their eye on a much longer term vision. Meeting young entrepreneurs like Klemen is inspirational as they are shaping foundations we can all build upon and this in one word is simply “greatness.”