I feel like this is especially relevant to me and I've thought about this topic a lot. I've known for a long time that MVs are now significantly more important than the quality of the music, which is why I spent much more money on the MVs for Millennia and Self-Destructive Girl. Of course, I still personally put a lot of effort into the songs themselves, but when I realized these songs had potential, I knew I would need MVs with gimmicks/GIFable moments. And what happened? Both songs got over 350,000 views when my previous average was like 20,000. I don't think these two songs are particularly much more amazing than my other songs, but it doesn't matter. The results speak for themselves.
In my opinion, MVs are about 75-80% of the success of a song now and can propel average songs to meteoric heights. I know that sounds pretty hyperbolic, but I honestly think I'm assessing it pretty accurately. This obviously mean that it prices out a lot of new producers since they can't afford a crazy MV. However, I wanted to give a bit of persprctivr using my own song analytics so new producers can gauge what they should do. With Millennia and Self-Destructive Girl popping off, I've made more money off just one of those songs than all my other songs combined. They've actually both made 2-3x the amount that it cost to make the MVs, whereas none of my previous tracks even came close to breaking even. If you're confident that you're song has the potential to pop off, it might be worth it to save up and invest more if possible.
To be clear, I really hate how the culture and trends have shifted since I'm a Vocaloid boomer and grew up listening to songs with single still images as the video. Unfortunately, I also feel like it's kinda just the times we live in and I've been forced to adapt.
Message to all casual vocaloid listeners: Please please please ask yourself whether you LISTEN to vocaloid music or or WATCH it. I don't judge if it's the latter, I just want you to think about it for a bit. I'm not forcing you but maybe you'd rather unlearn that