i was all 'man that doesn't look that good' and clicked around more visited https://nitter.net/RiversHaveWings which has a link to text2video visited https://huggingface.co/spaces/multimodalart/latentdiffusion which seems to be the current thing from there (takes a bit for it respond, doesn't show progress)
but y'know, i don't think the models have any idea what a 'cypherpunk' is supposed to look like, or have a strong sense of 'riding'. i'm a little surprised things haven't improved more at least for 'riding'. my ideal rule of thumb with synthetic stuff is that more time should be spent coding stuff, than using/generating it, unless i am generating code then i might bop it down to 30% of the time. don't really manage to pull that off.
