These are audio samples from the NU-Wave2 without overlaps. There are three audio files per folder: 
- wav.wav is the original audio in 48 kHz
- wav_l.wav is the downsampled audio at 16 kHz
- result.wav is the output of the model, attempting to reconstruct the upper frequencies

Each file has a corresponding image file with is spectrogram.
The audio is from the two folders (p280 and p315) of the VCTK that are not used during training/validation.