The first frame is the whole look of the video. The video model will move it, relight it a little and add motion blur, but it will not turn a glossy render into a phone photo. Everything that has to read as real is decided in this one still, which is why this is the longest lesson on prompting and why the frame is the cheapest place to be fussy: about thirteen cents and a few seconds per attempt, against dollars and minutes for every video take built on top of it.
By the end you will have written a 2,500 to 3,500 character still prompt from a template, generated it at the right settings, picked a keeper from a small batch, and written down what the keeper actually contains so the video paste can be written against it.