Poster still to a 5-second Wan 3.0 clip: cost at 480p, 720p, 1080p

A poster still from ChatGPT Image 2.5 at 1536x864 high costs about $0.0405, and a 5-second Wan 3.0 clip animated from it costs $0.3125 at 480p, $0.6250 at 720p or $1.25 at 1080p. The pair totals $0.3530, $0.6655 and $1.29.

The still is an estimate from the repository pricing code, and the Wan prices are the pricing_skus per-second rates in the Video API catalog: $0.0625, $0.125 and $0.25 per output second.

The handoff

Generate the still with POST /v1/images, then pass its Sume-hosted URL as a first_frame in frame_images on POST /v1/videos with model: "wan-3.0". Wan 3.0 accepts 2 to 30 seconds, ratios 16:9, 4:3, 1:1, 3:4 and 9:16, and both first_frame and last_frame.

If you send both frame_images and input_references, frame_images controls the mode and the request is processed as image-to-video, per the Video API docs.

{
  "model": "wan-3.0",
  "prompt": "Slow push-in on the poster, light moves across the paper",
  "frame_images": [
    {
      "type": "image_url",
      "image_url": {"url": "https://media.sume.com/img/EXAMPLE/0.png"},
      "frame_type": "first_frame"
    }
  ],
  "aspect_ratio": "16:9",
  "resolution": "720p",
  "duration": 5
}

Total by resolution

The still is a fixed cost, so the clip resolution is what moves the total.

Writing the motion prompt

The still sets the content, so the prompt should only describe motion. Name one camera move and one environmental change: a slow push-in, a light sweep, a drift of dust. Avoid telling the model to change what the poster says or add objects, because the first frame is fixed and a conflict makes the clip drift away from it.

Wan 3.0 also accepts last_frame, so a second still can define where the clip ends. That doubles the still cost but removes most of the guesswork about the final composition, which is useful when the ending is a logo or a product name.

• One camera move per clip.
• Keep the duration short while testing; the minimum is 2 seconds.
• Test at 480p, then rerun the approved prompt at 720p.

Choosing the resolution

Check motion at 480p first, because a bad idea costs $0.3125 instead of $1.25. Move up once the movement is right. Match the still's ratio to a ratio Wan lists, since 16:9 on the still and 16:9 on the clip avoids a crop.

A still re-used for several clips amortizes the image cost: four 5-second 720p clips from one poster cost $2.54 in all.

If your poster has text that must stay legible, test whether it survives the motion. Fine type is the first thing to blur or warp across frames, and a 1080p clip that mangles the headline is worse than a 480p clip that does not. Keep the headline large, keep camera moves slow, and check the final frame at full size. When the text is part of the brand, consider the overlay route from the compose post, where the type is added after the video exists and cannot be altered by the model.

For a quick quote to a client, the 720p row is the sensible default: it shows real motion and costs about a dollar less than 1080p. Offer 1080p as an upgrade once the concept is signed off.