pkgs/audio/voiceprint

Provides speaker embedding and voice identity detection abstraction, supports ECAPA, ERes2Net and NCNN-backed model paths, and maintains matching identity through vecid.

Go API References

Core structure and main function

Symbol	Function
Detector	Define audio input, speaker detection and lifecycle.
DetectorConfig	Configure model, threshold and vector identity behavior.
DetectResult	Returns speaker identity, confidence and embedding result.
DetectCallback	Receive detection events.
ConfidentGt	Express confidence threshold callback.
NewECAPA / NewERes2Net	Create the corresponding speaker model detector.

The Voiceprint package is responsible for signal feature, embedding model and identity detection, and does not own recording permissions, peer identity, user profile or biometric data retention policy; these are explicitly controlled by the calling product layer.