There’s a characteristic indie jankiness about the mix tonality here, with a significant upper-midrange spectral emphasis that gets increasingly grating the more the track’s texture builds. But I don’t think there’s any single root cause – it’s more like an unhappy confluence of several different contributing factors. The first of these is the sustained upper-register rhythm-guitar part first heard at 0:13, which has such a strong 3kHz presence peak in its response that any lead-vocal part would struggle to compete with it in terms of upfrontness (if that’s a word!), even were it not for the tidal wash of Bon Iver-esque feature reverb. I suspect that the fact that the vocal has then been actively distorted in this mix is partly a response to this guitar, the aim being to increase the vocal’s upper harmonic density, and thereby counteract the presence-region masking effects of the guitar’s dominant spectral peak. As a result, we get two different distorted parts pushing a lot of aggressive frequency energy into the 2-4kHz region, skewing the overall mix tonality towards harshness. It’s also plausible that further midrange boost might also have been applied at mastering in an attempt to aid lyric intelligibility, as the vocal balancing feels a little haphazard – the opening phrase of the choruses in particular feels really buried.

But there’s more. One of the problems with using distortion to emphasise vocal presence is that noise-based consonants tend to be over-emphasised in the process. As such, it’s usually necessary to do some extra processing and/or editing work to avoid this. But what it sounds like they’ve done here is primarily just de-ess the vocal to get the sibilance under control, but without similarly reining in the remaining noise consonants – in the chorus, for example, things like the ’t’ of “to being friends” or the ‘j’ and ‘sh’ of “just shared”. So in addition to the overall upper-midrange emphasis, we’re also getting flashes of more extreme HF abrasiveness from those unchecked consonants.

Finally, this track has been absolutely hammered into loudness processing, a key component of which is clearly digital clipping, judging by the copious flat-topping of the mix waveform. Now, I have nothing against clipping as a loudness-maximisation tool in principle, but in practice I find that it tends to work best when the clipping threshold is set to catch only the waveform’s drum peaks, in which case the added distortion components will usually merge fairly benignly with the noisiness of each drum hit’s attack phase. In this track, however, the more steady-state sounds in the mix also seem to be hitting the clipper, adding an unwelcome sprinkling of continuous digital clipping distortion, which gets pretty fatiguing on the ear pretty quickly.