Wedding save-the-date video: photo to end card with Gemini Omni

For a save-the-date, send a photo of the couple as the first frame and a finished date card as the last frame, and let Gemini Omni fill the 8 seconds between them. On Sume that is frame_images with first_frame and last_frame on POST /v1/videos, and an 8-second 1080p clip costs $1.50. Design the last frame yourself so the names and date are exact; do not rely on the model to spell them.

Omni does generate readable text (Google's page says so, read 2026-10-07), but a card you made in a design tool is the only version you can proofread before you pay.

Why pin both ends

A text-only prompt gives you a stranger. A first frame fixes the people; a last frame fixes the typography and date. The Video Router doc lists image-to-video with an end image as a supported mode of the model, and Google's Omni and Veo 3.1 pages both describe first-and-last-frame generation. Because the text lives in an image you control, a typo is yours to catch, not the model's to invent.

• First frame: a sharp, front-lit photo at 16:9 or 9:16; Omni does not do 1:1.
• Last frame: the same aspect ratio, large type, high contrast, nothing near the edges.
• Both images must be public HTTPS URLs that Sume can download.

Prompt the in-between

Describe motion, not the endpoints. A good middle is 'the camera drifts back, petals fall across the frame, the scene softens into a cream-colored card'. Add the sound: 'soft piano, no dialogue'.

{
  "model": "gemini-omni-flash-1.1",
  "prompt": "The camera drifts slowly back from the couple, soft petals fall across the frame, the scene dissolves gently into a cream card. Audio: soft solo piano. No dialogue.",
  "duration": 8,
  "resolution": "1080p",
  "aspect_ratio": "9:16",
  "frame_images": [
    { "type": "image_url", "image_url": { "url": "https://example.com/couple.jpg" }, "frame_type": "first_frame" },
    { "type": "image_url", "image_url": { "url": "https://example.com/date-card.png" }, "frame_type": "last_frame" }
  ]
}

Cost and what the invite list changes

Sume bills the Video Router list rate times 1.25 per second, rounded up to the cent. The same 8-second job has different totals per resolution.

Honest limits

Sume does not list Veo 3.1, so the Veo rows are only a price reference for readers who already use Google's API. Veo 3.1 1080p accepts only 8-second clips per Google's page, which suits this use case but is not available on Sume. Omni gives you a vertical 9:16 option for story-style invites, and you can re-run the same two frames with a different middle prompt until the motion feels right. Check that faces stay recognisable at the first second before you share it widely.

Sending it to guests

Test the final file on a phone before sending. Vertical 9:16 plays full screen in messaging apps and stories; 16:9 suits email and a wedding website. Omni offers no square, so a feed crop needs an extra step. Keep the date card on screen for at least the last two seconds of the 8, so guests have time to read it, and place the date and names away from the screen edges, where apps draw their own interface. If a guest list is large, re-using the same two frames with a different middle prompt gives you variations without redoing the design work.