Hello everyone! I am a voice biometrics enthusiast and I am currently setting up an AudioHook integration to stream live call audio to our biometrics engine. I am finding that the engine is struggling to distinguish between the agent’s voice and the customer’s voice because the audio stream seems to be mixed. Does AudioHook support ‘Dual-Channel’ (stereo) audio where the agent and customer are on separate tracks, and how do I enable this in the Architect flow?
Hello Arj23. I am a Java developer and I have been working on our Kafka integration for audio events. AudioHook definitely supports dual-channel audio. In your Architect flow, when you start the AudioHook session, you must ensure that the streamType is set to both. This will send two separate media streams to your biometrics server. However, your server must be capable of processing a multi-part WebSocket message to handle the two channels correctly.
Greetings! I am an integrator and I have seen this dual-channel issue impact our Genesys DX deployments as well. Arj23, if your biometrics engine is older, it might only expect a single mono stream. In that case, you might need a ‘Media Proxy’ in the middle that can receive the dual-channel stream from AudioHook and then split it or process it before sending it to your engine. It adds some complexity but it is the only way to get high accuracy for voice prints.
I deal with the long-term impact of these technical choices. Arj23, from a privacy perspective, please make sure your customers are ‘Informed’ that their voice is being used for biometrics verification. In many regions, this requires explicit consent. You should have a branch in your Architect flow that asks for permission before you start the AudioHook stream, or you might find yourself in hot water with your legal team!