This is generated from the motion model fine tuned version that I trained last month.
It have been train with 200 something Kpop dance video clip.
And it have been ran for 8 times to get this model I feel like okay.
There are still background consistency need to be improve. Because the input training video clips are all came with flashy dancing light background. Maybe that affect.
update: here are some screenshot of the progress.




Creed
2024-05-10 00:17:30 +0000 UTCBenjamin Law
2024-05-06 08:39:56 +0000 UTCiMPEUS da Great
2024-05-06 08:31:39 +0000 UTCLex
2024-05-05 00:45:05 +0000 UTCLex
2024-05-05 00:43:56 +0000 UTCBenjamin Law
2024-05-04 19:43:09 +0000 UTCMarmotte
2024-05-04 19:41:08 +0000 UTC