Still plus clip on Sume: $0.36 to $2.08 a shot vs Vidu Q4 I2V

The short answer

Vidu Q4 Preview's image-to-video mode starts from a picture you already have. On Sume, which does not list Vidu, you can make that picture with an image model and then animate it with a video row through frame_images. A 5-second shot costs about $0.36 with a Seedream 5 Lite still and Wan 3.0 at 480p, and $2.075 with a 4K Nano Banana 2.1 still and Omni at 4K. The Vidu figures for the same 5 seconds run from $0.1575 at 540P with the 30% discount to $1.95 at 4K list.

Cost per 5-second shot

The still is paid once per shot and the clip is paid per second. The Vidu rows assume you already have the still, so they exclude its cost.

Sending the first frame

On POST /v1/videos, give the picture in frame_images with frame_type set to first_frame; on the Video Router, use the flat image_url. If you also send input_references, the docs say frame_images wins and the request runs as image-to-video. Pick a still with the aspect ratio you want in the video, because the clip will follow the frame.

The still costs the same whichever clip row you use: Nano Banana 2.1 is $0.10 at 1K and $0.20 at 4K, and Seedream 5 Lite is $0.04375. The clip cost depends on the row and resolution.

When to make the still first

Making the still first lets you reject a bad composition for a few cents instead of a few dollars. Five rejected Nano Banana 2.1 stills at 1K cost 5 x $0.10 = $0.50, less than one 4K Omni clip at $1.875. Approve the frame, then animate it once. The image docs list the image models and sizes, and the video generation docs show the frame_images shape.

Scaling a shot list

A 12-shot storyboard at the cheapest Sume pairing costs 12 x $0.3563 = $4.28. At the 1080p Omni pairing it is 12 x $1.0375 = $12.45. Vidu Q4 at 540P with the discount is 12 x $0.1575 = $1.89 for the clips alone. Add your own stills to compare fairly: 12 stills at the Seedream 5 Lite price of $0.04375 are $0.525, so Vidu plus Sume stills would be $2.415.