Wan 3.0 in ComfyUI: reference tags for images, video and audio
Official templates for text-to-video, image-to-video and reference-to-video, with clips up to 30 seconds and synced audio.
ComfyUI added official Wan 3.0 templates on 24 August 2026. Search 'Wan 3.0' in the Template Library and you'll find text-to-video, image-to-video and reference-to-video graphs. They run as Partner Nodes, so you need a Comfy account with credits. This is not a local open-weights run.
The reference-to-video node is the interesting one for consistency work. It accepts up to 10 reference images, 5 reference videos and 5 reference audio clips. You assign each one a job in the prompt with tags: @Image1 is the shoe, @Image2 is the character, @Video1 sets the motion, @Audio1 sets the voice or rhythm. References are numbered by type in the order you connect them, so wire them in a fixed order and keep a note of it.
According to the docs, Wan 3.0 generates up to 30 seconds in one pass with synced audio on by default, and the official repo also lists instruction-based video editing. Third-party spec comparisons put its ceiling at 1080p. Use Kling, Veo or Seedance when you need 4K delivery.
For multi-shot sequences, the boring method still works best: generate shot by shot, reuse the same reference set or the previous shot's last frame, and assemble in your editor.