Save-the-date video from two photos: Wan 3.0 first and last frame

Send two photos to wan-3.0, one as first_frame and one as last_frame, and it generates the 5 seconds between them for $0.32 at 480p, $0.63 at 720p or $1.25 at 1080p on Sume. For a save-the-date, that means the video opens on one real photo of the couple and ends on another.

The duration, the two frames and a one-line prompt are the entire request. Everything else, including the date, goes on as text afterwards.

How the two frames are sent

Each entry in frame_images needs a frame_type of first_frame or last_frame. Sume infers image-to-video from this field (video generation docs). Check supported_frame_images for the model in GET /v1/videos/models before you rely on a last frame, and send both frames when you want a defined ending.

{
  "model": "wan-3.0",
  "prompt": "The couple walks together, camera drifts forward, golden hour",
  "duration": 5,
  "resolution": "720p",
  "generate_audio": false,
  "frame_images": [
    {"type": "image_url", "image_url": {"url": "https://example.com/photo-a.jpg"}, "frame_type": "first_frame"},
    {"type": "image_url", "image_url": {"url": "https://example.com/photo-b.jpg"}, "frame_type": "last_frame"}
  ]
}

Price

Wan 3.0 lists at $0.05, $0.10 and $0.20 per second. Sume bills list times 1.25 and rounds up to the cent (video router docs).

Choosing the two photos

Pick photos with a similar background and light. When the two frames are very different, the model has to invent a transition and the faces are the first thing to shift. Look at the middle of each result, not just the ends.

Privacy and consent

The photos show real people. Only upload images you have the right to use, and share the finished video only with the people it features until they have approved it.

Practical notes

Put the date and names on screen as text in your editor, in a font you choose. A save-the-date is the case where a misspelled name is a real problem, and generated video text is not reliable enough for it.

If you want a version for each side of the family or for a different language, change only the overlay text. You pay for one generation, not for each version.