how to align to the target view?

[Fengzexu]

Hello author, I would like to know how to align the generated results to the target pose, is pose estimation required again? Why does the output surround video appear to rotate on multiple axes from the source image. Thank You in Advance！

[Fengzexu]

Especially zero123++, I find it difficult to align to the perspective of the input. Although the hybrid model directly aligns the output pose with the original image, it seems that its generation quality is not robust enough.