KeenASR SDK Release Notes

Release 2.1; November 20th, 2025

Feature cross-platform: we are introducing AudioQualityResult metric in the Response class, which is provided in the finalResponse callback. This class provides various metric that describe the quality of the audio signal.
Enhancement cross-platform: we’ve renamed AlternativePronunciation class to WordPronunciation and introduced isValid method that allows you to check the validity of the data before you pass it to the methods that create decoding graphs. Note that decoding graph creation will FAIL if you try to pass invalid WordPronunciation; this behavior is different from previous releases were those invalid pronunciations would be ignored.
Enhancement cross-platform: The Result object once again includes clean_text, the recognition output without special tokens such as <SPOKEN_NOISE>. (This was available in releases prior to v2.0)
Enhancement keenasr-web: We are releasing a browser-based developer demo that lets you experiment with the SDK via a graphical interface. It’s a great way to explore the API concepts hands-on.
Enhancement keenasr-web: Logging now uses Emscripten’s native logger methods, resulting in cleaner and more readable console output.

Enhancement keenasr-web: improved browser compatibility testing (Safari/Chrome) and updated documentation on browser support
Enhancement keenasr-web: (re)-expose word audio playing in the oral reading demo
Enhancement keenasr-web: make alternative pronunciation tag element optional in the developer demo
Bugs keenasr-web: fixed a bug with persistence of ASR Bundle in the local filesystem (avoides untar operation if bundle is already installed)

Feature keenasr-web is now in General Availability and on-par with SDK on iOS and Android
Enhancement Android SDK is recompiled to conform with Google Play requirements for 16KB page size on 64bit devices
Enhancement optimized computation of Goodness of Pronunciation scores for long response
Bugs fix for a bug in startListening when VADGating is turned on (cross-platform)

Bugs fix for a crash on some 32-bit Android devices when KASRResponse is finalized or SaveAudio method is called on the KASRResponse instance

Feature Dashboard uploads are now available in both iOS and Android SDKs via KASRUploader/KIOSUploader classes and queueForUpload method in KASRResponse/KIOSResponse classes
Enhancement on Android, KASRBundle API for copying assets does not require listing all the assets in the ASR Bundle directory
Bugs fix for StopListening not working properly when VAD gating is used
Bugs fix for a crash on older Android devices during computation of pronunciation scores in the final result
Bugs fix for a crash on some 32-bit Android devices when KASRResponse is finalized or SaveAudio method is called on the KASRResponse instance

Release 2.0 is a major upgrade across iOS and Android platforms including numerous new features and bug fixes.

Feature The KeenASR SDK provides language support for German and French, in addition to English (adults and children models) and Spanish.
Feature Support for goodness of pronunciation scoring for English langauge ASR Bundle
Feature Support for contextual decoding graphs
Feature Support for providing alternative pronunciations for decoding graphs. Alternative pronunciations can be used to model: a) pronunciations of words not in lexicon (including made up words), b) common mispronunciations
Feature Robust model for out-of-vocabulary words (“<SPOKEN_NOISE>” word)
Feature Support for control of <SPOKEN_NOISE> probability when creating decoding graphs, providing ways to be more or less strict with accents and mispronunciations
Feature Support for x86_64 ABI on Android (relevant for support of Chrome OS on Chromebooks with Intel CPU)
Enhancement Various optimizations for decoding graph creation time
Enhancement Decoupled json/audio file management from Dashboard uploads
Enhancement Consistent API and behavior across all supported platforms