How to reduce PoseNet inputResolution to make it perform faster

Hi! I’m using posenet and hence video capture, and not drawing the video image. My project uses node and want to track wrist positions of everyone who opens the browser.
However, the size of my video is the size of my canvas, and the larger my canvas is the laggier posenet becomes. I however do want my canvas to be the window width.

Is this pose lag because of my video resolution? Would reducing the resolution of the video make posenet less laggy? If so, how can do this? Is this with pixel density?

Thank you!

update: I tried reducing the video size but it only captured half of the canvas. I did find that posenet has options, such as imageScaleFactor and inputResolution: 256! I managed to reduce the imageScaleFactor, but am not sure why nothing shows up ie posenet does not work when I reduce the inputResolution by stating

ml5.poseNet(video,{inputResolution: 256}, modelReady)

… any explanation why it does that?