Wan 2.7 References to Video: Full Model Guide

Specifications at a glance

• Model ID: wan2.7-r2v (reference-to-video)
• Image input: reference images of your subject — sharp, well-lit, multiple angles recommended
• Prompt: up to 5,000 characters — reference subjects explicitly ("the character from the references")
• Output: 720P ($0.10/s list) and 1080P ($0.15/s list) + flat $0.05/s platform fee
• Aspect ratios: 16:9 and 9:16 — durations of 3–15 seconds
• Typical render time: 1–5 minutes

Identity lock at a lower price

Reference-to-video sits between text-to-video and image-to-video: the model extracts your subject's identity from the references and re-stages it in whatever scene your prompt describes — setting, action, camera and lighting remain fully promptable. That makes it the tool for series content: the same mascot in ten ads, the same character across an episodic story.

Wan 2.7 doubles down on direction: the 5,000-character prompt (versus HappyHorse's 2,500) leaves room to script multi-shot sequences around your consistent subject, shot by shot with explicit timing. Note that wan2.7-r2v outputs 16:9 and 9:16 — for square 1:1 reference videos, use HappyHorse r2v.

Pricing and how to select it

The list price is $0.10/second at 720P and $0.15/second at 1080P plus the flat $0.05/second platform fee — a 5-second 720P clip costs $0.75 versus $0.95 on HappyHorse. Public sharing saves 10% and subscription plans another 2–7%; the exact total is shown before you generate.

Open the reference images to video generator and select Wan 2.7 from the model dropdown at the top right (HappyHorse 1.1 is preselected). Your first 3-second reference-to-video clip is free on either engine.