The way that he's singing them does a lot of the work, but it's clear those inscriptions were made to be sung, not just spoken - and I'm almost sure that those verses are being split into "blocks" of four syllables each. Picking for example the first verse of the last inscription:

la¹ḥa²ga³ mut⁴ | wa¹laẓ² t̠a³rom⁴

I probably got the vowels wrong, but what matters is their placement. The bold is for the most obvious syllables, I think that they're long? It's clearly tailored to a rhythm.

You may have noticed the Paa’ letter. In semitic languages, there is the Pe letter, but when arabic diverged it became just the Faa’ letter.

That's a really common sound change across languages; [p] seems to lenite faster than other voiceless stops. Other languages showing the same changes are Old Japanese (that /p/ is nowadays /h/ [ɸ~ç~h] and Old Armenian (PIE *t *k *kʷ ended as [tʰ kʰ kʰ], but *p ended as [w] or [h]).

I highly recommend his channel, he has accurate readings and makes some incredible content.

I can confirm the accuracy based on this video - he got the 1500 Portuguese distinction between [s̻] and [s̪], the retracted [ɫ] and the unraised [e o] just right, even if no modern dialect AFAIK keeps all those features intact.