Two listing photos, one room transition: first and last frame

To make a room-to-room transition from two listing photos, send the first photo as image_url and the second as end_image_url to gemini-omni-flash-1.1 on Sume's Video Router. A 6-second clip costs $0.75 at 720p and $1.125 at 1080p.

The model invents the movement between the two frames, so it suits an open-concept space where one room flows into the next, not a transition across a closed door.

What Sume documents for first and last frame

The Video Router docs list image_to_video for this model with image_url and an optional end_image_url, 3 to 10 seconds, resolutions of 360p, 720p, 1080p and 4K, and 16:9 or 9:16. Native audio is always on. Google's launch post (read 2026-10-05) describes first and last frame control and scene extension up to 40 seconds in 10-second steps. Sume's documented envelope is 3 to 10 seconds per job, so chain jobs if you need longer.

A five-room listing

A five-room listing is four transitions: kitchen to living room, living room to hall, hall to bedroom, bedroom to bath. At 6 seconds and 720p that is $3.00 for 24 seconds of footage. At 1080p it is $4.50. Add a title card and a call to action and you have a 30-second tour for under $5 in generation.

Order the photos in the way a visitor would walk. The model has to imagine the path from one frame to the next, and a short, plausible path gives better results than a jump across the house.

One pair

Send one job per pair, each with its own Idempotency-Key.

curl -X POST https://api.sume.com/v1/video-router/generate \
  -H "Authorization: Bearer $SUME_API_KEY" \
  -H "Content-Type: application/json" \
  -H "Idempotency-Key: listing-88-kitchen-to-living" \
  -d '{
    "model": "gemini-omni-flash-1.1",
    "image_url": "https://example.com/kitchen.jpg",
    "end_image_url": "https://example.com/living.jpg",
    "prompt": "Smooth walk-through from kitchen to living room, steady camera, no people",
    "resolution": "720p",
    "duration": 6,
    "aspect_ratio": "16:9",
    "mode": "async"
  }'

Honesty and quality checks

Real estate has a rule that generated video must not misrepresent the property. Treat the clip as a visual bridge between real photos, not as a record of the layout. Do not let a model add a window, a view or a finish that the house does not have, and check each clip frame by frame against the photos before it goes into a listing. Where your local listing rules require it, disclose that the transitions are generated.

Check also that the photos have matching exposure and lens look. A wide-angle kitchen next to a tight bedroom shot will morph oddly. If a pair does not work, add a middle photo and make two shorter clips.

Price changes are better as captions, not new renders. Burn the price and the room labels onto the finished tour with a caption job, so an update costs a caption pass and not a render.

Length, resolution and sound

Think about what the clip is for. A listing page, a social post and an open-house reel all want a different length. For a social post, three to six seconds per transition is plenty, since the viewer scrolls. For a listing page, a 30-second tour of five rooms works, and the transitions should be slow enough to read the room. The docs' 3 to 10 second range covers all of that.

Resolution follows the placement. A social vertical at 720p is $0.125 a second, and an MLS or website tour at 1080p is $0.1875. Draft every pair at 360p first, at $0.0375 a second, and look at the motion before you pay for the final pass. A six-second 360p draft is $0.225, so a five-room draft set is under a dollar.

Last, decide about sound. Native audio is always on for this model and cannot be turned off at the request, so a silent tour needs the audio replaced or muted in assembly. A quiet music bed under real estate footage is common, and Sume's Timeline can lay one under the clips.

One last workflow tip: name the jobs by listing and room pair, such as the listing id followed by kitchen-to-living, and store the job id with the photo URLs you used. If the agent asks for a change to one transition a week later, you can re-run exactly that pair with the same photos, and the other clips in the tour stay as approved.