Ask Lemmy

That was interesting, it was quite a bit more boring than I expected, I know that sounds glib and immature but it's just when I hear about figures like that whipping people in to a frenzy I kind of assumed there'd be a bit more emotional appeal and a lot more peaks and valleys to the emotional affect. There's definitely times where I see it working, at the very beginning of the clip shortly after the original audio sample it seemed compelling, it's a bit more theory dense than I'd have expected but I guess I tend to forget that that was what he was selling, not just the warmongering he's famous for in English speaking countries.

I think this offered a bit of a window in to what it must have been like, but unfortunately the AI seemed to suffer a bit as time went on, especially accent wise. He started out sounding like a particular variety of English, as in from England in the UK, but with an oddly Australian lilt then briefly dipped in to just Australian without the English then a very long section of being an American which also corresponded to a change in the vocal quality to being more hoarse and broken. I don't know a lot about AI tools but I would wager this might have had to do with limited training data, maybe only that speech itself was used, in fact given the pretty short section at the beginning that said "original audio sample", maybe just that snippet was used to extrapolate the rest of the AI rendition of the rest of the transcribed and translated speech. That would explain why it seems so emotionally homogeneous throughout which probably lessens the charisma that's supposed to have been so famous. Judging by his physicality in that original sample I get the impression that even within the context of raving anger and self righteousness that in reality he imbued his speeches with more variety of tone than we're getting here. It feels like the AI had to do the best it could over a pretty long and dense text of the speech from an audio sample smaller than the resulting output, that might explain the meandering accent too. Also worthy of mention is the part where there's a particularly hard to parse and pretty long sentence that bafflingly leads directly in to a verbatim repetition of that exact same sentence, which definitely sounds like a glitch, I feel possibly like the confusingness of the sentence itself might perhaps be a translation issue as well.

An interesting aspect to me is how the tone and style of the speech, especially in the early section before things start going off the rails feel really reminiscent of an Australian politician called Malcolm Roberts and lo and behold if Hitler had to pick favourites from Australia's current political landscape, I think he'd be making his top ten.