I used the "improve prompt" option on my prompt "Girl in the "lollipop girl" reference image turns and walks to the sofa in the "living room" reference image and sits down", and got this:
"A soft-lit living room at sunset; the camera follows the young girl in a gentle tracking shot as she turns her back to the lens and walks gracefully toward a plush sofa. In one fluid motion, the camera pans around to capture her turning to face the camera before sitting down, framed in warm cinematic lighting with shallow depth of field."
And this is the video it generated: