Vidu Q4 takes 15 image refs: what Sume's reference limits are

Vidu Q4 Preview, launched on 2026-10-07, accepts up to 15 image references and up to 3 audio references. Sume does not list a Vidu model, so there is nothing to call. The reference limits Sume does document are 10 images and 3 video clips of at most 3 seconds each for Gemini Omni Flash 1.1, with no audio references, and fixed shapes for Genjutsu and H3 Max Recast.

What the Vidu release says

The launch release (read 2026-10-08) describes Vidu Q4 Preview as available on the Vidu web app and API, with up to 15 image references, up to 3 audio references, 10-bit color, output at 540p, 720p, 1080p, 2K and 4K, and pricing from $0.014 per second. That is all this page relies on; the release does not give the limits for video references.

Sume's current video catalog does not include a Vidu row, so a Sume request cannot select one. The catalog moves, so read GET /v1/videos/models rather than trusting any post.

Reference limits Sume documents

The table lists only limits stated in Sume's docs. Where the docs name a capability but no count, the cell says so.

How Omni references work

Gemini Omni Flash 1.1 is one catalog id that routes by the shape of the request. Send reference_image_urls (up to 10) and/or reference_video_urls (up to 3, each at most 3 seconds), and refer to the media in the prompt as <IMAGE_REF_0> and <VIDEO_REF_0>, zero-based in list order. It runs 3 to 10 seconds at 360p, 720p, 1080p or 4K in 16:9 or 9:16, with native audio always on.

Because there is no reference_audio_urls field, you cannot pass a voice or a track as a reference. If you need sound you control, mute the clip path or finish in Timeline.

• Billing is per output second by resolution: a 10-second 720p clip is $1.25 as of 2026-10-08.
• An Omni request with references still must stay within 3 to 10 seconds.

Choosing for a reference-heavy brief

Count your references first. If the brief needs 12 product angles, Omni's 10-image cap forces you to choose ten. If it needs a spoken reference, Omni has no field for it.

Do not infer a Vidu price from the release: $0.014 a second is a starting price on Vidu's own platform, and Sume does not resell it. For Sume, price the same 10 seconds at 720p on Omni: $1.25.

What this does not say

The Vidu release is a launch announcement and gives limits and a starting price, not a test. Sume does not list a Vidu model, so there is no Vidu price or route to compare here. The only like-for-like items are the reference counts, and those come from each side's own page. Whether 15 image references give better results than 10 is not something either page states.

If your brief is reference-heavy, the practical question is how many distinct references you really have. Ten images and three short clips cover most of a product shoot; more than that is usually a sign that the references should be merged into a board first.