text to 3d soon, via a so-called neural radiance field. Basically gradient descent and minimise loss from random viewing angles of 2d renderings.
Maybe gunship anon could use this as a sort of 3d mesh DALL-E for his quests? These models are probably all static though but it is a start.
You can load the 3d models on the website here.
There is a frog.
https://dreamfusion3d.github.io/ 