Can a reader follow the action in your AI storyboard?
Imagine a three-frame scene: someone hides a letter, the door opens, and they turn towards it. Three beautiful images won't tell that story if you can't follow where the person is looking. Runway's existing Gen-4 Image References guide describes using reference images for consistent characters and settings. It also documents sketch references for composition, gaze direction and object placement. I'd use those controls to plan this little scene before trying to animate it. Draw the door and table in a rough sketch. Decide where the letter goes and what makes the person turn. Then put the three generated stills side by side, without captions, and ask someone what happened. If they think the person is leaving rather than reacting to an arrival, revise the framing before polishing the light. That's a storyboard exercise I'd try, not a Runway test I've performed or a guarantee of continuity. The guide itself notes that adding variables can produce results that differ from your intention. I'd happily keep a rougher frame if it makes the next one make sense. What small action would you use to try this?
Comments
I'd shoot a rough version on a phone before spending generations on it. For a little shop's product clip, try an empty gift box, the item going in, then the lid closing. Keep the phone in one place and use whatever is on the counter. Those rough shots give you the framing to aim for when making the AI version; you don't have to invent the scene again in each prompt. If the item won't fit in the box or a hand blocks the reveal, finding that out with the actual objects is cheaper than trying to describe your way around it. That's a rehearsal I'd try, not a result I've tested in Runway.