Improved Real-Time Transcription Speed and Accuracy

We're excited to launch an update to our Real-Time WebSocket Transcription API! This update improves both accuracy and latency for results streamed back from the API.

Improved Real-Time Transcription Speed and Accuracy

Accuracy Update

Over the past few months, we've seen a huge uptick in developers looking to implement Real-Time Transcription into their applications and products. Real-Time Transcription is powering innovative accessibility features, like closed captioning of online events, live coaching of customer support and sales calls to help improve the customer's experience, and a slew of other interesting use cases and applications.

At AssemblyAI, we're focused on building the easiest to implement, and most accurate API for automatic speech recognition. That's why we have a great team of Speech Scientists and Deep Learning Engineers focused on rapidly improving the performance and accuracy of our models - and today we're excited to be releasing an improvement to our Real-Time WebSocket Transcription API, that returns more accurate and faster results for developers.

New Helper Libraries

As part of this release, we're also working on launching more sample code and helper libraries to implement our WebSocket API. A great example of that is our new demo, that shows how to stream audio from the browser to our WebSocket API using JavaScript and WebRTC. All the code for this demo is open source! So you can easily use it as inspiration for your own applications.

The above GIF shows a glimpse of just how fast and accurate our real-time API can stream transcription results back!

Wrapping Up

For more information about our Real-Time WebSocket API, you can check out the public API Docs. You can also write to us any time at, or at @AssemblyAI on Twitter!