RNNoise model assets are intentionally absent from the default SoundSavo 0.26.0 QA archive.
To build a model-enabled local QA candidate, extract the COMPLETE PROJECT and run:

  npm run neural:prepare
  npm run verify
  npm run package

The build-time fetch downloads an EXACTLY PINNED third-party package and verifies SHA-256.
No neural assets are downloaded while watching browser videos.
If assets are absent, the DSP path continues to work normally.
