X-Unimotion Support: Possibly a better SteadyDancer?
https://byteaigc.github.io/X-Unimotion/
It seems to be doing what SteadyDancer is doing, but also keeps facial expressions. One of the biggest issues with wan animate is setup for characters of different body proportions and or starting composition.
This seems to solve the issues in theory, per the examples. It is also provided by Bytedance which is a pretty big company. 😦
One could run x-unimotion as a standalone perhaps, and feed the transposed output into WAN Animate, perchance.
🤞🤞🤞
that looks pretty good ;-) (chance of it being cherry picked, but looks promising for sure) It doesnt look like its any released models or code though, but maybe its upcoming
( And looks like its part of a multi-research https://zeyuan-chen.com/X-Dancer/ being another one, that apparently just needs music input and image. (but not sure what it is, or if its even wan ;-)) )
This model has not been released.
that looks pretty good ;-) (chance of it being cherry picked, but looks promising for sure) It doesnt look like its any released models or code though, but maybe its upcoming
( And looks like its part of a multi-research https://zeyuan-chen.com/X-Dancer/ being another one, that apparently just needs music input and image. (but not sure what it is, or if its even wan ;-)) )
This has not been released either.
Yes, i noticed it was not out yet. And X-Unimotion was announced all the way back in July with "Will be released very soon". So who knows if it will be, could just be research, or abandoned for better things
(but looked really good in the examples, hopefully something will come)
Yea mb, I was on my phone and assumed it was released. But I did email them on their page and tweet replied to one of the devs in his post about it so let's see😂
I also emailed them back in July but never got a reply. It looks like a promising project, but maybe it did not perform as expected and they just abandoned it.
https://byteaigc.github.io/X-Unimotion/
It seems to be doing what SteadyDancer is doing, but also keeps facial expressions. One of the biggest issues with wan animate is setup for characters of different body proportions and or starting composition.
This seems to solve the issues in theory, per the examples. It is also provided by Bytedance which is a pretty big company. 😦
One could run x-unimotion as a standalone perhaps, and feed the transposed output into WAN Animate, perchance.
🤞🤞🤞
being sniping it since july, i feel they never gonna release it, it's like they planned to release it and they figure out it's too powerful so they can't deliver it to the community, it's always like this.