Unveiling the Power of devAIce® SDK 3.13 and devAIce® Web API 4.6: Exploring the Newest Features  

,
Milenko Saponja

We are excited to announce a major release of the devAIce® SDK version 3.13, as well as devAIce® Web API 4.6, a milestone packed with new features and improvements designed to help developers build even better audio applications. 
 

Multi language support for Automatic Speech Recognition (ASR) 

The Automatic Speech Recognition (ASR) module has been enhanced with multi-language support. Previously, English was the only supported language, but recognizing the diverse needs of our clients, we’ve now integrated support for 99 languages, powered by the underlying whisper.cpp library (see full list of supported languages). 
Since our models are language-independent, integrating multi-language capabilities into the ASR aligns with our strategy of breaking down language barriers. This upgrade ensures global support, making devAIce® an even more powerful all-in-one audio analysis tool. 
The module can auto detect spoken language and also user can specify the certain language if this information is given which can boost the accuracy even further. 

Enhanced Speaker Verification Module with support for multiple speakers 

The Speaker Verification module now supports verifying multiple speakers, making it much easier to analyze conversations involving several people simultaneously. 
The module usage remains the same, consisting of two steps: the Enrollment step and the Verification step. The key difference in this version is that, during the Enrollment step, you can now enroll N speakers, and in the Verification step, you can verify against all previously enrolled speakers. 
This enhancement reduces the number of required verification runs and improves overall performance.  
To help users get started with the updated module, we’ve added both Python and C++ code samples demonstrating its usage. 

Renaming of Emotion to Expression 

AI cannot detect how humans feel, but rather how we express ourselves. It identifies emotion-related behaviors by analyzing patterns in visible expressions, such as facial movements, gestures, and vocal cues. As a result, we have renamed the term “Emotion” to “Expression.” The previously named Emotion modules – Emotion (Large), Emotion, and Multi-Modal Emotion – are now called Expression (Large), Expression, and Multi-Modal Expression. The same applies to all underlying functions and variables, which have introduced breaking changes. 
This renaming is a precautionary step to ensure that the terminology more accurately reflects the output of the modules and to prevent any potential misunderstandings. 

In addition to these changes, several other improvements and bug fixes have been made. As always, we encourage all devAIce® users to update to the latest version to take full advantage of these exciting enhancements. 

Further information on our Voice AI technology and the use cases we offer for customised solutions can be found at audEERING®.

related posts

No data was found