Have you tried just going image-to-depth map? You could render the image in 3d with shallow depth then just shift the camera a bit
The main issue with moving mouths is the relative stillness, it's enough to introduce coherent motion in the user's field of vision.
If anything it might give you a unique distinction vs d-id, the head movements it doesn't aren't natural at all, but it's enough to not make it feel stiff
Interesting. I guess camera movement on 3d scenes would make the video more dynamic. That will put the focus more the "video aspect" as opposed to "Avatar" aspect of the product. I might explore that feature a little bit.
The main issue with moving mouths is the relative stillness, it's enough to introduce coherent motion in the user's field of vision.
If anything it might give you a unique distinction vs d-id, the head movements it doesn't aren't natural at all, but it's enough to not make it feel stiff