Metadata-Version: 2.4
Name: streamsad
Version: 0.1.0
Summary:  A light-weight, streaming speech activity detection (SAD) module, designed for real-time speech-only classification
Author-email: Mohammad Hassan Sohan Ajini <mohamad.hasan.sohan.ajini@gmail.com>
Project-URL: Homepage, https://github.com/mohamad-hasan-sohan-ajini/streamsad
Project-URL: Repository, https://github.com/mohamad-hasan-sohan-ajini/streamsad
Project-URL: Issues, https://github.com/mohamad-hasan-sohan-ajini/streamsad/issues
Requires-Python: >=3.7
Description-Content-Type: text/markdown
License-File: LICENSE
Requires-Dist: numpy
Requires-Dist: onnxruntime
Dynamic: license-file

# streamsad

`streamsad` is a streaming-oriented Speech Activity Detection (SAD) module that operates frame by frame, without requiring access to the full audio signal (unlike batch processing). Unlike simple energy-based Voice Activity Detection (VAD), it accurately detects human speech while ignoring music, background noise, and silence. Powered by an efficient ONNX model and a postprocessing algorithm inspired by WebRTC (using ring buffer smoothing), it runs entirely on the CPU with minimal overhead, making it ideal for real-time voice interfaces, ASR frontends, and low-resource deployments.
