Corridor Digital is really pushing their new AI film-to-anime video:
https://www.youtube.com/watch?v=FQ6z90MuURM
This time they're using something called WarpFusion that's supposed to make Stable Diffusion’s output more temporally consistent across frames.
But it doesn't do a very good job; the result is still painful to watch due to temporal inconsistency.
Seems like there’s been a ton of research in related areas lately, but not quite this use case:
https://github.com/showlab/Awesome-Video-Diffusion
comex /
npub1nk…3nfnz
2023-08-13 21:06:54
Author Public Key
npub1nkv2kkavdkh8pc53dvwf6ta9efm3pqy8g5h63p68xl56c252c4pq43nfnzSeen on
wss://relay.noswhere.comPublished at
2023-08-13 21:06:54Event JSON
{
"id": "7da3d67f73067d3a2aef1aa10965ffacfafd261a4d7f361e52b58af14ebe56d2",
"pubkey": "9d98ab5bac6dae70e2916b1c9d2fa5ca77108087452fa8874737e9ac2a8ac542",
"created_at": 1691960814,
"kind": 1,
"tags": [
[
"proxy",
"https://mas.to/users/comex/statuses/110884343916365246",
"activitypub"
]
],
"content": "Corridor Digital is really pushing their new AI film-to-anime video:\nhttps://www.youtube.com/watch?v=FQ6z90MuURM\nThis time they're using something called WarpFusion that's supposed to make Stable Diffusion’s output more temporally consistent across frames.\n\nBut it doesn't do a very good job; the result is still painful to watch due to temporal inconsistency.\n\nSeems like there’s been a ton of research in related areas lately, but not quite this use case:\nhttps://github.com/showlab/Awesome-Video-Diffusion",
"sig": "522ea7ad83ca89d49d55d5828fe355ecf40c5797ab1cc17c2ec88c31727a75b51f78ab3ca4cea178e4ae3191342a58ae12d28444b77e96b9d962b836076dde75"
}