Wan 3.0 Prime Spicy ref2v: Ten References, One Bill

alibaba/wan-3.0-ref2v-prime-spicy takes three kinds of reference in one request: up to 10 images, up to 5 videos, up to 5 audio clips. Most multi-reference endpoints charge you per reference. This one does not. Its billing formula counts video seconds and nothing else, which changes how you should assemble a reference set.

Key Takeaways

• The billing formula printed on the console Pricing block is billable duration = input video duration (up to 5 seconds) + output video duration. Images and audio do not appear in it.
• Ten reference images cost the same as one: $0 extra. The reference video is the metered input, and it stops metering at 5 seconds no matter how long the clip is.
• MiniMax H3 ref2v prices the other way, adding $0.40 for every reference image past the fifth. Same job, very different bill.

What the Endpoint Actually Accepts

Required fields are model, prompt, duration, size, and aspect_ratio. The reference arrays are all optional: images (max 10), videos (max 5), audios (max 5). duration is any whole second from 2 to 30. size is 480p, 720p, or 1080p. aspect_ratio offers 16:9, 9:16, 1:1, 4:3, 3:4, and adaptive.seed accepts -1 to 2147483647. prompt_expansion_enable defaults to true, so the endpoint rewrites your prompt unless you set it false — turn it off while you are still deciding which references do the work. There is no negative_prompt, the same as every other Wan 3.0 endpoint.

Only Video Is Metered

The Pricing block states the rule directly: both input and output videos are billed by duration in seconds, and billable duration = input video duration (up to 5 seconds) + output video duration. Nothing in that formula scales with image count.The cap cuts both ways. A 12-second reference clip bills 5 seconds; a 3-second clip bills 3. Trimming a long reference below 5 seconds saves nothing, so hand over the whole clip and let the cap absorb it.

What a Real Job Costs

Prices below come from the console model-detail Pricing block on 2026-09-17. The tier carries a 20% off badge, putting the charged 480p rate at $0.060 per billable second.

The middle row is the one to internalize. Adding a 12-second reference video to a 10-second job costs $0.30, because only the first 5 seconds meter. Adding nine more reference images to any row costs nothing.That is the opposite of how MiniMax H3 ref2v bills. Its Pricing table has a Reference Images Count column: the first five images are included, and each image after that adds $0.40 regardless of resolution. Nine images on a 5-second 768p job is $2.00 there against $0.40 for five. If your workflow leans on large reference sets, that difference decides the endpoint before anything about output quality does.The plain tier behaves identically on billing — see Wan 3.0 Spicy ref2v for the character-consistency angle, and Wan 2.7 Uncensored vs Wan 3.0 Spicy ref2v for the generation gap. Siray exposes all of them behind one key, so moving between them is a change to the model string.

Developer Notes

Build the reference set first, then pick the endpoint the set is cheap on. The discount badges move; re-read the Pricing block the day you run a long batch.Create your free Siray account and send ten references in a single ref2v call.

A Note on Compliance

Siray's NSFW and uncensored models generate legal, consensual adult content only — CSAM and any illegal content are refused with zero tolerance. "Spicy" means no moderation beyond what the law requires, not that anything goes. Wan 3.0 Prime Spicy is a third-party Alibaba Cloud model available through Siray's API, not a Siray-hosted model, and every reference image or video you upload must be material you have the right to use.