With Zep's RLE compression thingy you can now store an absurd amount of graphical data. This means short FMVs are technically possible, if unwieldy.

I don't know what to use it for besides converting porn GIFs to Pico-8's palette and playing them over an energetic chiptune backbeat, though?

Something like Tecmo theater maybe?

I am imagining something on this scale is all that's possible. 64x64. No dithering. A single tone scale. Few frames. The less motion the better, of course. With transparency, GIF-style frame-combining optimizations can be done as well, since you don't have to clear the screen in between decompressing the next frame and whatnot.

But because it's fun to play with Pico-8's palette and dream of bigger things, here's a full size "what if" version.

conventional idea: attract mode

less so: interactive multimedia gif art (i like this a lot, thinking about what you could do with this excites me)

Oi! I don't quite get the specifics regarding the limitations here, but I'm working on a game that features a similar approach to animating, maybe that'd lend itself to this rather fittingly?

... Pico-8 doesn't have that many blues, though?

The limitations here are mostly "whatever compressed data we can fit into the RAM somewhere", whether that takes up part of the sprite data, map data, or an absurd count of script characters in a giant string. Plus the Pico-8's palette and screen sizes, of course. This is very variable, so we don't really have a strict idea of exactly how many frames of animation and how complex the animation can be is possible yet. :P