Camera orbit and zoom transition prompts for Omni first and last frame
Google says Omni 1.1 Flash handles orbits, zoom transitions and loops between two keyframes. Prompt patterns and costs for the Sume API, from $0.19 per draft.

For a camera orbit or a zoom transition, send two keyframes that frame the subject from different angles or distances and name the camera move in the prompt. Google lists complex camera orbits, zoom transitions and looping clips as use cases of Omni 1.1 Flash first and last frame control (Google blog, read 2026-10-05); on Sume the model id is gemini-omni-flash-1.1.
A 5-second orbit draft at 360p is $0.19 on Sume, and the 1080p final is $0.94.
Choose frames that imply the move
The two frames carry the camera move. An orbit needs the same subject in the same place with the camera at two positions around it, for example front view and three-quarter view. A zoom transition needs the same framing at two scales, for example a wide street shot and a close-up of one shopfront. If the two frames show different subjects, the model has to invent a cut, and the prompt should say that it is a transition between scenes instead.
- Orbit: same subject, same height, camera moved 45 to 90 degrees around it.
- Push-in or pull-out: same center, different scale.
- Loop: first and last frame the same image, with a prompt that describes a motion that returns to its start.
- Scene-to-scene zoom: the end of the zoom in frame one is the start of the next scene in frame two.
Prompt patterns
Keep the prompt about the middle of the clip. Use camera verbs the model can follow: orbit, dolly in, pull back, whip pan, rack focus. State direction (clockwise, left to right), speed (slow, steady) and what stays still (the subject). Avoid describing the first or last frame again, since the images already do.
- "Slow clockwise orbit around the lamp, the background drifts, the lamp stays centered."
- "Steady push-in from the street to the shop window, ending on the sign."
- "Seamless loop: the plant sways and settles back to where it began."
Costs
Orbits read well at 5 seconds; slower moves like a long push-in may need 8 or 10. The price grows linearly with seconds, rounded up per clip, on top of Sume's 1.25 multiple on the list rate (Video Router page).
| Seconds | 360p | 720p | 1080p |
|---|---|---|---|
| 3 | $0.12 | $0.38 | $0.57 |
| 5 | $0.19 | $0.63 | $0.94 |
| 8 | $0.30 | $1.00 | $1.50 |
| 10 | $0.38 | $1.25 | $1.88 |
Why this is a frame problem first
Camera moves in two-keyframe video are mostly decided by the pair of images. The prompt can say "orbit", but if the second frame shows the same angle as the first, nothing in the pair asks for an orbit and the model has little to travel toward. Build the pair so that it could not be anything else. For a 90 degree orbit, shoot the front and the side. For a push-in, shoot the wide and the tight.
Sume does not limit what the two frames are beyond the model's envelope: 16:9 or 9:16, public image URLs, one first frame and one last frame, with the clip 3 to 10 seconds. Everything else is composition.
Troubleshooting the move
If the orbit wobbles, the frames probably disagree on subject position. Re-shoot or re-crop so the subject sits at the same point in both. If the move is too fast, lengthen the clip from 5 to 8 seconds: the same change happens over more frames. If the subject drifts, add "the subject stays centered and does not move" to the prompt.
For loops, test by playing the clip twice end to end. A visible jump at the seam means the last frame differs from the first, or the motion does not return to its start.
Draft first
Orbits and zooms are cheap to try at 360p: $0.19 per 5-second try against $0.94 at 1080p. Ten drafts cost $1.90, which is about two 1080p tries. Pick the frame pair and the prompt that read best, and spend the final on that one.
Request
Send the frames as frame_images to POST /v1/videos, with first_frame and last_frame entries, and set the same aspect ratio for the crops: 16:9 or 9:16.
{
"model": "gemini-omni-flash-1.1",
"prompt": "Slow clockwise orbit around the lamp, background drifts, lamp stays centered",
"frame_images": [
{"type": "image_url", "image_url": {"url": "https://example.com/lamp-front.png"}, "frame_type": "first_frame"},
{"type": "image_url", "image_url": {"url": "https://example.com/lamp-side.png"}, "frame_type": "last_frame"}
],
"resolution": "360p",
"duration": 5,
"aspect_ratio": "16:9"
}Sources
Related posts
More in Media tools
- Can a 1080x1080 square video be a 3-minute YouTube Short?
Yes: YouTube's help page says a Short can run up to 3 minutes if it is square or vertical. Render 1080x1080 in Sume Timeline 1.0 for $0.30 at 180 seconds.
- Remove the card behind burned-in captions: colors.card null
Set design.colors.card to null on a Sume caption job to show no card. It works on styles that support design, not on punch or tiktok-green.
- Captions with no style set: Korean gets black-outline, no gold tint
If you send no style, Sume picks one from the caption text, and the spoken word gets no gold tint. Set design.colors.active to add one.
- Cheapest product video ad: six stills in a Timeline render for $0.30
Six product photos become a 15 s vertical ad with no video model: a $0.10 Timeline render plus $0.20 caption cues. $0.30 against $3.43 for five Wan clips.
Written by Sume