This is what I initially thought, however, I have already witnessed working demoes of 3DGS when using a single viewpoint, but armed with additional auxiliary data that is contextual relevant to the subject.