Renovation reveal: Omni first frame before photo, last frame after

Turn a before photo and an after photo of a room into a 10-second Gemini Omni 1.1 Flash reveal on Sume with image_url and end_image_url. Request and cost math.

4 min readSume
All posts

To make a renovation reveal from two photos, send the before photo as image_url and the after photo as end_image_url in one gemini-omni-flash-1.1 request on Sume. The model generates the 3 to 10 seconds between them. Google lists first and last frame control as a feature of Omni 1.1 Flash (Google, read 2026-10-05), and Sume's Video Router takes the pair as an image-to-video request (Sume docs: Video Router, read 2026-10-05).

Shoot the two photos so they can connect

The model has to invent the path from one frame to the other, so the pair has to share a camera position. Use these rules.

  • Same spot, same height, same lens. A tripod mark on the floor does the job.
  • Same aspect ratio. Set aspect_ratio to 16:9 or 9:16 to match both photos.
  • Same light, or the clip has to invent a change of light as well as a change of room.
  • Clear the clutter in the before photo only if it was cleared in the after photo too, or say in the prompt what moves.

The request

Draft at 360p, 6 seconds, and look at the result before you commit to a finish.

curl -X POST https://api.sume.com/v1/videos \
  -H "Authorization: Bearer $SUME_API_KEY" \
  -H "Idempotency-Key: kitchen-reveal-draft-1" \
  -H "Content-Type: application/json" \
  -d '{
    "model": "gemini-omni-flash-1.1",
    "prompt": "Locked camera. Cabinets, counters and lighting transform from the old kitchen to the finished one.",
    "image_url": "https://media.sume.com/artifacts/artf_demo/before.jpg",
    "end_image_url": "https://media.sume.com/artifacts/artf_demo/after.jpg",
    "resolution": "360p",
    "duration": 6,
    "aspect_ratio": "16:9"
  }'

Cost of the pass

The rates are the example provider rates of $0.03 and $0.10 per second times 1.25. They are not a quote; confirm them against the catalog and the pricing page before you rely on them. A failed job is refunded.

Worked cost for a 6-second reveal at the example rates, which are placeholders to replace with GET /v1/videos/models (read 2026-10-05)
StepSecondsExample Sume rate per secondCost
Draft at 360p6$0.0375$0.225
Finish at 720p6$0.125$0.75
Both12n/a$0.975

When the middle looks wrong

A reveal over a big change can show warped furniture mid-clip. Shorten the clip, since fewer seconds leave less room to invent, or change the prompt to name the transition: a wipe of light, a camera push, a dust-sheet pull. If the two photos differ in camera angle, retake one of them; no prompt makes up for a moved camera.

Keep the 360p draft as the approval file. Send the finish only after the client or your team says yes.

Think about audio too. Native audio is always on for this model on Sume, so the clip comes with generated sound. A silent room reveal can come out with ambient noise or music you did not ask for. If the clip will go under a voiceover or a music track, describe the sound you want in the prompt, such as quiet room tone, and mute or replace it in the timeline.

For a full house, make one reveal per room and join them. Four rooms at 6 seconds make a 24-second tour, one billable timeline minute, and every room is a separate draft you can approve or re-roll on its own. This also keeps a bad kitchen from costing you the good living room.

Keep the photo pair with the clip in your records. If the owner later asks what the room looked like before, the first frame is already there, and if a clip is questioned, you can show the two real photos it was built from.

Say in the listing, caption or page that the middle of the clip is generated. The before and after are real photographs; the transition between them is not a record of the work.

Sources

Related posts

More in Use cases

All Use cases posts

Written by Sume