Further Advances in AI Video

The latest in AI music video developments

Now that AI video is starting to catch up to the excellent state of AI music, the obvious question is what can be done in terms of combining the two. In last week’s post, I showed two early efforts in that space.

Only two days later, a new direct-by-vibe interface dropped, along with access to both Seedance 2 and Seedance Mini 2, both of which offer more accurate character consistency, higher-resolution video up to 4k, better lip-syncing, and more realistic visuals, all of which can be seen in this hard-hitting Soulsigma cover of a Vibe Patrol original, Float Down the River.

Believe it or not, the original is very ethereal, with an almost The Orb-like vibe and a female singer. That’s not so much the case here.

The new Vibe Director system is incredibly effective, but it costs 10x more in token terms than the previous one, and that’s just at the 720p level. This is still pretty cheap in real-dollar terms: that video cost about $35 to make from start to finish, but it’s considerably more than the $4 or so that the previous two both cost. I should mention that I haven’t finished Chiba City Blues yet, because testing out this new OpenArt system took priority.

To move up to 4k video would be cost-prohibitive at the moment. That would run around $250, which is still astonishingly inexpensive for that level of HD video, but is rather more than I can justify dropping on a single experiment at this point, especially when the 720p results are so good.

But if we can pick up a few more paid subscribers here, perhaps we can start working on some more detailed experiments, especially some comparative ones. We already have more than 80 original songs from which to choose, although some of the older ones definitely need to be reworked to bring up their audio quality to current levels.