How should a resolution selector map to the documented Vidu size field?

A CometAPI Vidu resolution selector should submit an exact dimension string, not its human-facing tier label. In CometAPI’s Vidu Q3 guide, the API parameter section lists size as an optional string and recommends 1280x720.

For the documented Vidu route, the guide gives this mapping:

Display label based on the guide’s tier names Value to serialize as size
540p tier — 960 × 528 960x528
720p tier — 1280 × 720 1280x720
1080p tier — 1920 × 1080 1920x1080

The first row deserves care: the guide calls it a 540p tier and describes it as 16:9, but 960x528 is neither 540 pixels high nor exactly 16:9. Keep the documented string; don’t “correct” it to 960x540. Showing the dimensions alongside the tier makes that distinction visible when someone is planning a composition.

Keep labels out of the request value

My recommendation is to maintain a small mapping for this CometAPI route and reject unrecognized selections before constructing a request. This is a local serialization fragment, not a complete API call:

const viduSizes = new Map([
  ["540p", "960x528"],
  ["720p", "1280x720"],
  ["1080p", "1920x1080"],
]);

function appendViduSize(form, selectedTier) {
  const size = viduSizes.get(selectedTier);
  if (!size) throw new Error("Choose a listed Vidu size");
  form.set("size", size);
}

Here 720p is an application-side key only; the serialized field is 1280x720. Test all three mappings and an unknown selection locally. Avoid silently falling back to a different tier: the user should know which dimensions they selected. If the form starts at 1280x720, make that an explicit UI choice—the guide’s recommendation does not establish the server’s default when size is omitted.

Bound the promise to this route

This is a documentation-based mapping, not a live-tested claim that the API rejects 720p or accepts every request carrying these strings. The guide was published August 27 and the source snapshot was fetched September 7, 2026. It points to the Vidu endpoint documentation for current fields and limits and warns against assuming provider-native options work through CometAPI’s unified endpoint.

Before extending this selector to portrait or square output, check the current route documentation rather than extrapolating from these three values. The list alone also tells us nothing about automatic cropping or the quality of a generated clip. For a render form, exact dimensions are the useful promise here—not an inferred framing or quality guarantee.

The exact size value is also worth retaining as a comparison key in a render-cost report, rather than grouping jobs only under “HD.” But pixel count is not a billing model: 1920x1080 contains 2.25 times as many pixels per frame as 1280x720; that arithmetic establishes neither a 2.25× price nor a 2.25× generation time.

For a bounded comparison, I would hold the model, requested duration, prompt set, and other supported settings constant, then record the submitted size, observed completion time, actual charge, and whether each clip meets a predefined delivery requirement. Include charged unsuccessful attempts in total spend and divide by the number of usable clips—not just completed jobs. If none qualify, report that outcome rather than a cost-per-usable-clip figure.

That separates “which dimensions did we request?” from “which setting is economical for this deliverable?” A lower-resolution option is not automatically cheaper per usable clip if it requires more attempts or extra finishing work. Until current route pricing and measurements are available, I would leave savings estimates out of the selector; the documented size list alone cannot support them.