Consistent Character Animation! (AI news)

[Legolas18]

As some of you already know, I have jumped on the AI train 😁

So, here is an upcoming tech (it hasn't been released yet, but an early 2024 release is probable) that can also be used by DAZ users ..and that's animations from a picture! (for more info, go here)

1) IMAGE2VIDEO:

Online Demo: MagicAnimate (similar project)
https://huggingface.co/spaces/zcxu-eric/magicanimate

You basically start with any picture (so it could be a photo, a Daz character, or an illustration), and you need an existing video (the AI extracts the OpenPose skeleton on each frame), while maintaining character and clothing consistency!

Video 1
Video 2

NOTE: I can already see people releasing "bones" animations to be used for various projects.

2) ADD ANY CLOTHES TO PHOTO: OutfitAnyone

Online Demo: 
https://huggingface.co/spaces/HumanAIGC/OutfitAnyone

But that's not all! There is also the possibility to change clothes on a picture! And then animate it too!

Video 3

Let's hope eventually that they add environmental animated effects, camera control and all that stuff 😋

3) TALKING HEADS:

This seems still a bit meh, but it's getting there:

And, of course, you can upscale your video to 16x and even multiply frames by 16x by using a program like Topaz Video AI 4

[gts6]

@legolas18 That looks great, looking forward to it. That site has some comparisons of theirs to "other". That implies theres already some ways available to do this, do you know of any currently available methods to do this?

[Legolas18]

@gts6 I found a Video2video:

https://github.com/volotat/SD-CN-Animation

as well as a talking heads:

https://github.com/OpenTalker/SadTalker

As for image to animation, I didn't find anything good yet.

[Ethiopia]

This could be seriously fun. One could have great sport with various ladies,...tee hee....

[gts6]

@legolas18 thanks, yeah im already familiar with that one. I havent seen animating an image yet, either.

I do know there is one that exists that Google has, I remember reading about it last year. It can take a starting image and an ending image frame, and then fill in the blanks between with AI generated frames to complete the video. Wish I could remember the name of it.

[Legolas18]

Added links to the online demos 🙂

[Author]

Posts