sense-sdk-cpp-tutorials

• sense-file/ - file-mode: load a local audio file and run inference to completion (sense-file-app <audio_file>).
• sense-stream/ - stream-mode: capture the microphone and infer in real time (sense-stream-app).

1. Build

Prerequisites: a C++14 compiler, CMake >= 3.10, and the Cochl.Sense Edge SDK staged at ./sense (include/ + lib/).

cmake -B build
cmake --build build

The binaries sense-file-app and sense-stream-app land in this directory, next to config.json and audio-files/, so you can run them in place.

2. Run

Set your project key in kProjectKey (top of sense_file.cc / sense_stream.cc). A config.json in the working directory is used if present (missing is non-fatal).

./sense-file-app audio-files/babycry.wav     # file mode (blocks until end of file)
./sense-stream-app                           # mic mode (Ctrl-C to stop)

Each prints one JSON result per inference window (start_time, end_time, prediction_time_ms, tags); a summaries field is added when Result Summary is enabled.

3. Notes

• config.json (optional) controls sensitivity, Result Summary, Audio Activity Detection, and Automatic Gain Control (AAD/AGC are stream-only); the same knobs are available at runtime via processor->controls() after Build().
• Callback: each tutorial's ResultPrinter (a sense::FrameResultListener) fires OnResult once per window on the SDK's inference thread -- edit it or pass your own to .SetCallback().
• Capture spec (stream), at the top of sense_stream.cc: kNumChannels, kFormat, kCaptureSampleRate (0 = model rate; higher is downsampled), kHopSize. kFormat is sense::SampleFormat::{kFloat32,kInt16,kInt32}; feed 24-bit audio as kInt32, left-justified (sample << 8). sense-file reads its format from the file header. Microphone capture is tutorial-side (vendored miniaudio); the SDK has no audio-device dependency.