Abstract: Methods and apparatus are disclosed for generating predictions of tendencies of a waveform segment to induce frisson. In one example, a method includes producing Mel-frequency cepstral coefficients corresponding to an audio waveform, dividing the waveform into a plurality of segments that have been tagged with a value indicating a likelihood that the segment will produce a frisson response in a listener, and training a neural network with the Mel-frequency cepstral coefficients and the segment tag values to generate predictions of a tendency of a waveform segment to induce frisson. In some examples, the method further includes displaying a visualization of the waveform, wherein the visualization indicates how likely portions of the visualized waveform are to induce frisson.