Issues

• With several target phrases, a shorter phrase is generated as an adversarial negative for a longer one

  dscripka/openWakeWord#354 · Filter by author wilbowes opened on Oct 3, 2026

• Default threshold 0.5 costs most detections in noise; 0.25 measured at 83% vs 48%, with no false positives

  dscripka/openWakeWord#350 · Filter by author sacha9214 opened on Sep 17, 2026

• Further benchmark results

  dscripka/openWakeWord#349 · Filter by author DanBmh opened on Sep 17, 2026

• Licence of the shared feature models (melspectrogram, embedding_model) for a commercial product

  dscripka/openWakeWord#348 · Filter by author rmunoz928 opened on Sep 16, 2026
  1

• License of melspectrogram.tflite and embedding_model.tflite for a commercial app

  dscripka/openWakeWord#346 · Filter by author Compasse-Shelly opened on Sep 7, 2026

• [Help] 16KB issue with old tensorflow lib

  dscripka/openWakeWord#340 · Filter by author Ayush0Chaudhary opened on Jul 6, 2026
  3

• License clarification: Are melspectrogram.onnx and embedding_model.onnx under Apache 2.0 and suitable for commercial use?

  dscripka/openWakeWord#338 · Filter by author RellyLee opened on Jun 25, 2026
  1

• ONNX inference backend produces near-zero scores on macOS ARM64 (Apple Silicon)

  dscripka/openWakeWord#336 · Filter by author wenjielei1990 opened on Jun 21, 2026
  5

• Following the documented augmentation recipe yields very high false-positives/hour (~177/hr); RCA points to background-mix on positives collapsing positive/negative separability.

  dscripka/openWakeWord#335 · Filter by author karthikeyareddy-create opened on Jun 20, 2026

• Google Colab notebook needs update

  dscripka/openWakeWord#331 · Filter by author BillBBBB opened on May 23, 2026
  12

• README outdated

  dscripka/openWakeWord#330 · Filter by author NovaH00 opened on Apr 13, 2026

• Brother seriously add a statement that you have website where we can know that you have a website to train/download custom models

  dscripka/openWakeWord#329 · Filter by author Ayush0Chaudhary opened on Apr 12, 2026
  1

• no support for Browser level wake word detection - HTML and JS

  dscripka/openWakeWord#328 · Filter by author Vedapani0402 opened on Apr 8, 2026

• Bug - demo not working

  dscripka/openWakeWord#323 · Filter by author Beenhakker opened on Mar 18, 2026

• cannot install current version on raspberry pi

  dscripka/openWakeWord#322 · Filter by author hslr4 opened on Mar 16, 2026
  2

• Training and inference script for multiclass wakeword detection

  dscripka/openWakeWord#321 · Filter by author kiran-infiswift opened on Feb 25, 2026
  1

• can I replace the melspectrom model with a melspectrom extractor like kaldi-native-features?

  dscripka/openWakeWord#320 · Filter by author wbjnpu opened on Feb 25, 2026

• Training custom models in 2026: dependency fixes + Dockerized and Native solution

  dscripka/openWakeWord#317 · Filter by author briankelley opened on Feb 4, 2026
  4

• Gradient Accumulation is wrong

  dscripka/openWakeWord#316 · Filter by author qZhang88 opened on Jan 29, 2026

• stream mel spectrogram error

  dscripka/openWakeWord#315 · Filter by author qZhang88 opened on Jan 28, 2026

• Model License Question

  dscripka/openWakeWord#313 · Filter by author AleksanderB31 opened on Jan 23, 2026

• creating issue with Nabu remote connect

  dscripka/openWakeWord#312 · Filter by author osokthedevil opened on Jan 15, 2026

• ❓ Questions / Help / Support

  dscripka/openWakeWord#310 · Filter by author bil21071 opened on Jan 14, 2026

• macOS (Apple Silicon) cannot run openWakeWord natively due to missing TFLite runtime

  dscripka/openWakeWord#309 · Filter by author goppi0 opened on Jan 13, 2026
  3

• In train.py, convert_to_tflite is always attempted after training even when it's not specified on the command line

  dscripka/openWakeWord#305 · Filter by author gtjoseph opened on Jan 8, 2026