svoid trainVAD() {
  final BWImage img1 = frequencyImage(voiceMegaMix());
  final BWImage img2 = frequencyImage(nonVoiceMegaMix());
  int clipLength = iround(2.0*spectro_pixelsPerSecond());
  int stepSize = clipLength/4;
  print(+clipLength);
  L<BWImage> images1 = map(func(IntRange r) -> BWImage { bwHorizontalClip(img1, r) }, stepIntRange(clipLength, intRange(0, img1.getWidth()), stepSize));
  L<BWImage> images2 = map(func(IntRange r) -> BWImage { bwHorizontalClip(img2, r) }, stepIntRange(clipLength, intRange(0, img2.getWidth()), stepSize));
  print(allImageSizes(concatLists(images1, images2)));
  print(l(images1) + " + " + l(images2) + " images");
  
  LPair<BWImage, Bool> trainingList = trueFalseBPairs(images1, images2);
  
  replace Recognizer with F1<BWImage, Bool>.
  
  new Best<Recognizer> best;
  scorePreciseRecognizer(best, func(BWImage img) -> bool { false }, trainingList);
  scorePreciseRecognizer(best, func(BWImage img) -> bool { true }, trainingList);
}