Image to drone shot AI: turn a photo into an aerial move

To turn an image into a drone shot with AI, send the photo to an image-to-video model as the clip's first frame and describe an aerial camera move in the prompt, such as rising over the scene, flying forward, or orbiting a subject. The model generates the flight from that frame on, so whatever lies beyond the photo's edges is invented. The result is a generated shot, not footage of the real place.

Facts come from Sume's Video generation, Video Router, and Media inputs docs and the Sume API reference, read on 2026-09-28. Anything described as current behavior is read from Sume's API code.

How do I turn a photo into a drone shot?

Without code, describe the shot in the Agents tab: the agent picks the models and asks before it spends. Over the API:

• Pick a photo with room to fly: open sky, a skyline, a coastline, or a building seen from above. Crop it to the shape you want, such as 16:9.
• Put it at a public HTTPS URL. Localhost, private-network, non-HTTPS, and signed or private URLs are refused.
• Send it to POST /v1/videos in frame_images as the first_frame, and write the move in prompt.
• Set an aspect_ratio and a duration the model lists, and set resolution explicitly.

curl -X POST https://api.sume.com/v1/videos \
  -H "Authorization: Bearer $SUME_API_KEY" \
  -H "Content-Type: application/json" \
  -H "Idempotency-Key: drone-coast-001" \
  -d '{
    "model": "seedance-2",
    "prompt": "Aerial drone shot: the camera rises slowly from the beach and flies forward over the cliffs toward the lighthouse, golden hour light",
    "frame_images": [
      { "type": "image_url", "image_url": { "url": "https://example.com/coast.jpg" }, "frame_type": "first_frame" }
    ],
    "aspect_ratio": "16:9",
    "resolution": "720p",
    "duration": 8
  }'

What should an AI drone shot prompt say?

The request has no camera field, so the move goes in the prompt. Sume's docs suggest details about motion, camera angles, lighting, and scene composition, and camera movement prompts covers the general vocabulary. Name one move, its direction, and its speed:

• Rise and reveal: "The camera rises slowly from the rooftop and tilts down, revealing the whole city at dusk."
• Fly forward: "A low aerial shot flying over the waves toward the lighthouse."
• Orbit: "The camera circles the house at roof height, keeping it in the center of the frame."
• Pull back: "The camera flies backward and up until the whole island fits in the frame."

Can I make a drone shot without a photo?

Yes: send only a prompt that describes both the place and the move. Every model in the table below except grok-imagine-video-1.5 runs from text alone; that one needs a first frame. With a photo, you can also pin where the flight ends: add a last_frame next to the first_frame on a model that takes one, as in Earth zoom out AI video.

Will the drone shot come out in the shape I ask for?

It depends on the model. Each model lists the aspect_ratio values it accepts, and in current code a value it doesn't list is refused; Vertical 9:16 video generation API has the lists. With a first frame, though, current code passes aspect_ratio on to only some models, so crop the photo to the shape you want before you send it:

What are the limits?

• The surroundings are invented. A generated drone shot is not a record of a real property or piece of land; Real estate photo to video AI covers what that means for listings.
• The prompt steers the move, and no field sets it. Nothing promises the model follows it, so watch the clip.
• One clip runs 30 seconds at most, on seedance-2.5 or wan-3.0; AI video length limits by model lists the rest. For a longer flight, start the next clip on the last frame of the one before.
• Each clip is billed by model, at the rates GET /v1/videos/models lists in pricing_skus.