☁️BizyAir Wan2.6 Image To Video

Wan 2.6 shipped in December 2025, commercial-only, no weights - the release where the wider community's read on Alibaba's Wan series shifted from "waiting for the next open version" to accepting that the open era of the numbered Wan line had likely ended at 2.2. This node is the only route to 2.6 from ComfyUI as a result: there's no local checkpoint to load, because there isn't one.

It sits right next to the Wan 2.5 node in this pack and shares most of its shape - same audio capability the open Wan checkpoints lack entirely, same prompt-rewriting toggle - but with a couple of real differences worth knowing before you pick one over the other.

What's actually different from Wan 2.5 here

• Longer clips. 2.5 caps out at 10 seconds; 2.6 offers 5, 10, or 15.
• No 480P option. 2.5 offers three resolution tiers; 2.6 only offers 720P and 1080P. If you specifically wanted the cheaper/faster low-res tier, it's not available on this node.
• A shot_type control. This is genuinely new, not present on the 2.5 node at all - it picks whether the generated video reads as one continuous shot or as multiple cut-together shots, per the node's own tooltip. Real creative control, and one you'd otherwise have to fake by generating and manually splicing multiple separate clips.

Inputs and outputs that matter

• image - required, the starting frame.
• audio (optional) - same as 2.5, feed your own track and the video syncs to it.
• prompt / negative_prompt (optional).
• resolution - 720P or 1080P, defaulting to 1080P.
• duration - 5, 10, or 15 seconds.
• prompt_extend (default true) - same LLM-rewriting behavior as the 2.5 node: helps short prompts, adds latency, worth turning off once your prompt is already detailed.
• shot_type (default single) - single or multi. Per the tooltip, this only takes effect when prompt_extend is true - if you've turned prompt rewriting off to save time or preserve your exact wording, this setting silently stops doing anything. Worth knowing before you conclude the control doesn't work.
• auto_audio (default true) - auto-generates audio when you haven't supplied your own; your audio input takes priority if present.

Outputs: video, actual_prompt (what actually ran through generation after any rewriting), and bizyair_model_name.

How to install it

• ComfyUI Manager - search "BizyAir", install, restart.
• Manual - cd ComfyUI/custom_nodes && git clone https://github.com/siliconflow/BizyAir.git, restart.
• Comfy CLI - comfy node install bizyair.

Set your API key first - the in-app login prompt or bizyair.siliconflow.cn. There's no local fallback for this model; every call is billed through your BizyAir credits.

Common issues & troubleshooting

shot_type doesn't seem to change anything. Check prompt_extend - it has to be true for shot_type to have any effect at all. This is the single most likely reason the setting appears to do nothing.

Generated video content doesn't match your literal prompt. Same story as the 2.5 node: prompt_extend may have rewritten it. Check actual_prompt to see what actually went into generation before assuming the model ignored you.

You wanted a cheaper 480P run and can't find it. Not available on this node - 2.6 only offers 720P and 1080P. If cost matters more than the resolution bump, the 2.5 node's 480P tier is the one to use instead.

No audio despite feeding one in. Confirm the audio input is actually connected rather than left empty, and check auto_audio isn't somehow being relied on by mistake - supplied audio should always take precedence per the node's documented behavior.