Get started · Lesson 5 of 7
How to write AI video prompts like a film director
Call the shot size, one camera move and the light, one shot per prompt. The four words that turn a vague clip into a scene.
You need: One scene in mind. Nothing to upload. · Start to finish, step 5 of 10
Step by step
Call the shot size first
Wide, medium or close-up. The shot size decides what the viewer is allowed to notice: a wide shot makes the weather bigger than the person, a close-up makes one raindrop matter.
Name one camera move, only one
Slow push-in, tracking shot, orbit or crane up. Each move means something: a push-in says the moment is getting serious. Two moves in one prompt fight each other.
Describe the light
Overcast storm light is flat and cold, golden hour warms the same cliff, neon makes it a city at night. Light is the fastest way to change the mood without changing the scene.
Write one shot per prompt
One subject, one action, one move. Want a sequence? Make each shot on its own, then cut them together in Film Studio or Merge Videos.
Put it in order
Shot size, subject and action, camera move, light: "Close-up of an old lighthouse keeper, silver braids under her yellow hood, rain on her cheeks, slow push-in on her eyes as she spots a light at sea, cold overcast storm light."
Change one thing, compare
Make the wide shot and the close-up of the same keeper, then watch them side by side. The wide makes the storm bigger than her; the close-up makes one look matter. Same scene, different emphasis.
Why it works
- Models follow the first words hardest. Lead with the shot size and the subject, and leave adjectives for the end.
- Camera words are a shared language: every video model on PixelDojo understands push-in, tracking, orbit and crane.
Made this way
Your practice task
Write the same scene three ways: wide with a crane up, medium with a tracking shot, close-up with a push-in. Compare the clips.
Read the transcript(1:38)
Write video prompts like a director using shot size, one camera move, and light to shape the same lighthouse scene. Start your Seedance 2.5 prompt with the shot size, wide shot, then name the keeper her silver braids and yellow raincoat with waves exploding below the cliff beside a white lighthouse. Give the camera just one move, a slow crane up, finish with golden hour light breaking through the clouds. That gives the shot its movement and lighting without adding another scene. Then press generate. While the wide shot renders, write the next prompt. Start with close-up of the same keeper. Keep her silver braids and yellow hood but bring attention to rain on her cheeks. Now call a slow push in on her eyes as she spots a light at sea. End with cold overcast storm light. Keep it to one shot per prompt and one camera move per shot. Once both takes are ready, compare what each shot emphasizes. The wide shot makes the storm bigger than her. The crashing water and rocky coast give the scene its sense of scale. The close-up makes one look matter instead. Her face becomes the focus rather than the coastline. To change the emphasis, change the shot size, the move, and the light, not the whole paragraph. You've called two shots with different emphasis. Next, try another camera move in PixelDojo while keeping your subject consistent.