Company

About Soniox

Soniox mission is to accelerate the adoption of speech-based applications and spark innovation of human-machine voice interaction. We have developed the most accurate speech recognition system and made it freely available for anyone to use*.

Soniox builds artificial intelligence and infrastructure for audio and sound understanding.

Research

We have developed a novel approach to building speech recognition systems on unlabeled data. We have applied this approach to English language and built a system with near error-free speech recognition for most audio data.

The Soniox speech system is significantly more accurate than any other leading speech recognition system in the industry. The near error-free recognition dramatically improves downstream applications that require speech as input. It also opens doors to new kinds of applications and human-machine voice interaction.

World of Languages

For massive adoption of speech recognition in the world, it is critical to have an approach that easily scales to many if not most languages. This is one of our biggest strength, since we can apply the same approach to many other languages.

Our short-term goal is to near-error free recognize top 20 languages.Our long-term goal is to accurately recognize most of the spoken languages, dialects and accents in the world.

Engineering

Infrastructure is an equally important pillar for Soniox. It has to support all of the requirements that come with the audio and sound AI processing systems. We have built several libraries to support real-time low-latency computation over the entire stack. Our platform automatically scales to adjust to the current load of incoming audio streams, and can run on any cluster or cloud environment with commodity CPU machines.

From Ideas To Applications

Developers are at the forefront of Soniox. We simplified speech integration by designing new and more useful interfaces for speech recognition. With a few lines of code, you can transcribe live audio from microphone, local files or any other stream in real-time low-latency scenarios. With Soniox Web Speech, you can integrate speech recognition into the website or web application. We have significantly reduced the iteration cycle to build, test and deploy speech based applications.

Engineering

Infrastructure is an equally important pillar for Soniox. It has to support all of the requirements that come with the audio and sound AI processing systems. We have built several libraries to support real-time low-latency computation over the entire stack. Our platform automatically scales to adjust to the current load of incoming audio streams, and can run on any cluster or cloud environment with commodity CPU machines.

From Ideas To Applications

Developers are at the forefront of Soniox. We simplified speech integration by designing new and more useful interfaces for speech recognition. With a few lines of code, you can transcribe live audio from microphone, local files or any other stream in real-time low-latency scenarios. With Soniox Web Speech, you can integrate speech recognition into the website or web application. We have significantly reduced the iteration cycle to build, test and deploy speech based applications.


Speech Everywhere

We strongly believe that the speech recognition and voice-based interaction era has only started. And, that speech recognition as a service should be ubiquitous and accessible to any developer and user. These are the main reasons we decided to offer the most accurate speech recognition service in the industry for free* .

We hope that our contribution will accelerate the adoption of speech-based applications and spark innovation of human-machine voice interaction.