So, this works frame by frame using SD, extracting a pose (for example) and generating a new frame using the prompt and the pose, then re-encoding gen'd frames

So, this works frame by frame using SD, extracting a pose (for example) and generating a new frame using the prompt and the pose, then re-encoding gen'd frames as a video? Seems interesting. Any demos?