VDK Runtime 6.3.0 - 2nd June 2026

Summary

We’re excited to announce the release of VDK Runtime 6.3.0, introducing major enhancements across voice recognition, speech enhancement, session management, and system-level performance improvements.

This release brings Neural ASR with Conformer models, expanded audio buffer capabilities, improved session pipeline validation, and significant upgrades to VDK-Service, including SSL support, new audio routing capabilities, and improved lifecycle control for audio components.

What's New in VDK Runtime 6.3.0?

Voice Recognition

This release introduces a major upgrade to voice recognition capabilities, improving both accuracy and flexibility in real-world conditions.

Introduced a new neural-based recognition engine, delivering more accurate and robust speech recognition
Enhanced system behavior for better responsiveness during live recognition sessions
Improved error reporting to make issues easier to understand and resolve
Expanded support across core recognition APIs for more consistent performance

Security & Connectivity (SSL Support)

VDK Runtime 6.3.0 adds secure communication support to make deployments safer and more production-ready.

Added support for SSL encryption for HTTPS and WebSocket (WSS) connections
Services can now automatically run in secure mode when certificates are provided at startup
Introduced secure startup support on Android to simplify protected deployments

Learn More

Additional updates and refinements have been made across the VDK ecosystem to enhance developer experience and performance.

We’re committed to improving usability, flexibility, and performance — and your feedback on VDK 6.3.0 is always welcome.

Versions

Vdk-Service

Component	Version
Vdk-Service	2.1.0 Updated

Samples (C++)

Component	Package name	Version	Technologies
Chained grammars	chained-grammars	7.1.0 Updated	Voice Recognition (Grammar)
Dynamic grammar	dynamic-grammar	7.1.0 Updated	Voice Recognition (grammar)
Simple application	simple-application	7.1.0 Updated	Voice Recognition (grammar) Voice synthesis
Tts	tts	7.1.0 Updated	Voice synthesis
Voice biometrics	voice-biometrics	7.1.0 Updated	Voice Biometrics
Voice Commands Language Understanding	voice-commands-language-understanding	4.1.0 Updated	Voice Recognition (continuous)
Speech Enhancement	speech-enhancement	3.1.0 Updated	Voice Recognition (graMmar) SPeech enhancement
Free Speech	free-speech	4.0.0 Updated	Voice Recognition (NEURAL)

VSDK (C++)

Component	Version
Vsdk	11.1.0 Updated
Vsdk-csdk-asr	2.1.0 Updated
Vsdk-vec-addon	2.1.0 Updated
Vsdk-csdk-tts	2.1.0 Updated
Vsdk-idvoice	6.1.0 Updated
Vsdk-tssv	5.1.0 Updated
Vsdk-vnlu	4.1.0 Updated
Vsdk-s2c	3.1.0 Updated
Vsdk-audio-portaudio	5.1.0 Updated

Samples (JAVA)

Component	Package name	Version	Technologies
Chained Grammars	chained-grammars	7.4.0 Updated	Voice Recognition (Grammar)
Dynamic Grammar	dynamic-grammar	6.4.0 Updated	Voice Recognition (grammar)
Simple Application	simple-application	6.4.0 Updated	Voice Recognition (grammar) Voice synthesis
Speech Enhancement	speech-enhancement	3.4.0 Updated	Voice Recognition (graMmar) SPeech enhancement
Tts	tts	7.4.0 Updated	Voice synthesis
Voice Biometrics	voice-biometrics	4.3.0 Updated	Voice Biometrics
Free Speech	free-speech	5.0.0 Updated	Voice Recognition (NEURAL)

VSDK (JAVA)

Component	Version
Vsdk	7.4.1 Updated
Vsdk-csdk-asr	4.1.0 Updated
Vsdk-vec-addon	2.1.0 Updated
Vsdk-csdk-tts	4.1.0 Updated
Vsdk-idvoice	4.4.0 Updated
Vsdk-tssv	4.4.0 Updated
Vsdk-vnlu	1.1.0
Vsdk-s2c	4.1.0 Updated

Details

VSDK (C++)

Vsdk

Features & Improvements:

Voice Recognition
- Added a new neural model type using Conformer models instead of acmod.
- A Recognizer now supports either: multiple acmod, or a single conformer.
Speech Enhancement
- Added new fields to SpeechEnhancerInfo:
  - inputChannelCount
  - referenceChannelCount
  - outputChannelCount
  - aecEnabled
  - supportedSampleRates
Buffer
- Added support for creating buffers from files or streams (.wav and raw PCM formats).
- Added support for saving buffers to files or streams as WAV or raw PCM formats.
- Deprecated saveToFile(): Use saveToWav() or saveToRawPcm() instead.
File Consumer
- Added support for saving audio as WAV files based on file extension, with fallback to raw PCM format.
File Producer
- Fixed invalid delays between audio buffers.
- Added support for loading WAV files based on file extension, with fallback to raw PCM format.
Pipeline
- Updated producer getter behavior: When requesting a producer with a custom type, the getter now returns nullptr if the type does not match.

VSDK (Android)

Vsdk

Features & Improvements:

Voice Recognition
- Added new error codes for improved diagnostics and error handling.
Voice Synthesis
- Fix invalid EventCode mapping.
- Added new error codes for improved diagnostics and error handling.
Voice Biometrics
- Fix idvoice model storage structure (Each model is now stored in a dedicated folder for better organization and management).
Buffer
- Added support for retrieving audio duration in seconds.
- Added channel extraction from audio buffers.
- Added support for appending partial byte or short arrays to buffers.
- Added support for creating buffers from files or streams (.wav and raw PCM formats).
- Added support for saving buffers to files or streams as WAV or raw PCM formats.

VDK-Service

Vdk-Service

Features & Improvements:

Security & Connectivity
- Added SSL support for HTTPS and WebSocket (wss) connections. If certificates are provided at startup, the service automatically runs in secure mode.
- Added a new Android method: startWithSsl() for securely starting the service.
Voice Recognition
- Added support for updating model_settings in VoiceRecognition consumers while the pipeline is running.
- Added support for Neural ASR in session VoiceRecognition modules.
- Added support for Neural ASR in the following routes:
  - /v1/advanced-recognition/recognize
  - /v1/voice-recognition/recognize
Voice Synthesis
- Session updates for VoiceSynthesis consumers are now only allowed when the pipeline is stopped.
Audio Recorder Support

Added support for using the audio recorder in the following routes:
- /v1/advanced-recognition/recognize
- /v1/voice-recognition/recognize
- /v1/voice-recognition/userwords/enroll
- /v1/voice-biometrics/authenticate
- /v1/voice-biometrics/identify
- /v1/speech-enhancement/enhance
Audio Player Support
Added support for using the audio player in the following routes:
- /v1/voice-synthesis/synthesize
- /v1/speech-enhancement/enhance
Audio Player
- Added audio player progress events: AudioPlayerProgress.
Session Pipeline
- Added pipeline integrity checks during session startup:
  - Sample rate and channel count compatibility are now validated before startup.
  - The pipeline will fail to start if inconsistencies are detected.
- Added modifier index support to resolve ordering issues when multiple modifiers are present:
  - Lower index values are processed first.
  - Duplicate index values are not allowed.
  - The index remains optional when only one modifier is used.
Speech Enhancement
Updated the response format of /v1/speech-enhancement/enhancers:
- Enhancer details are now returned as objects keyed by enhancer name.
- The previous enhancers key remains available but is now deprecated and will be removed in a future release.
- Added new enhancer information fields:
  - input_channel_count
  - reference_channel_count
  - output_channel_count
  - aec_enabled
  - supported_sample_rates
Android
- Fixed an issue preventing vdk-service from restarting after being stopped on Android.
- Added a new vdk-service AAR package containing:
  - A service manager class for starting and stopping vdk-service.
  - Integrated HTTP and WebSocket clients for interacting with supported routes.
API Fixes
- Fixed DELETE /v1/voice-recognition/userwords/:user and DELETE /v1/voice-recognition/userwords
  These endpoints now correctly use query parameters instead of request bodies.