Wan 3.0 reference limits on Sume: 10 images, 5 videos, 5 audio clips

Wan 3.0 (wan-3.0) on Sume accepts up to 10 reference images, up to 5 reference videos with a combined length of 15 seconds and a frame rate of at least 16 fps, and up to 5 reference audio clips with a combined length of 15 seconds. Those limits come from the Video Router catalog constraints in the Sume repo, as of 2026-10-08.

Wan 3.0 is one of the more flexible reference models in the catalog, and the limits differ from MiniMax H3 and Seedance, so they are worth reading before you assemble a request.

How to send references

On POST /v1/videos, send references in input_references: image entries use image_url, and the model accepts a reference type only if supported_input_references lists it. On the Video Router, use the flat fields reference_image_urls, reference_video_urls and reference_audio_urls. Only public HTTPS URLs work.

If you send both frame_images and input_references on /v1/videos, frame_images controls the mode and Sume processes the request as image-to-video, so do not mix them when you want reference-to-video.

curl -X POST https://api.sume.com/v1/video-router/generate \
  -H "Authorization: Bearer $SUME_API_KEY" \
  -H "Content-Type: application/json" \
  -d '{"model":"wan-3.0","prompt":"The character from the images walks to camera","reference_image_urls":["https://example.com/a.png","https://example.com/b.png"],"resolution":"720p","duration":6,"aspect_ratio":"9:16"}'

What Sume does not expose

The catalog notes that file_url, web_url and enable_thinking are not exposed in v1 for Wan 3.0, and that it has no bitrate_mode. If your client is written against another provider's Wan schema, drop those fields or Sume will reject the request.

Cost does not change with references

Wan 3.0 is billed by output second at the resolution you pick: $0.0625, $0.125 and $0.25 per second at 480p, 720p and 1080p (provider list $0.05, $0.10, $0.20, times 1.25). The reference count does not add a charge in the catalog, unlike MiniMax H3, which lists each reference image after the first five at $0.08 provider list. A 6-second 9:16 clip at 720p with eight references is 6 x $0.125 = $0.75.

Practical limits

These limits are listed as catalog constraints, so validate counts, total durations and frame rates on your side before you submit; a request that breaks them is a wasted round trip at best.

• Keep the reference videos under 15 seconds in total, even if you send fewer than five.
• Use a 16 fps or higher source; lower frame rates break the constraint.
• Wan 3.0 aspect ratios are auto, adaptive, 16:9, 4:3, 1:1, 3:4 and 9:16.
• Check GET /v1/video-router/models for the live capabilities before you hard-code limits.