XB_WanDanceSwitcher

This is the simplest node in the whole pack, and it knows it. XB_WanDanceSwitcher isn't a sampler or a VAE - it's a prompt switch for the Wan Dancer music-to-dance workflow. Pick a dance style and a motion amplitude, and it hands you the two text strings that drive the dance generation.

That's the entire job, and it's genuinely handy. Wan-Dancer is Alibaba's Apache 2.0 music-to-dance model (released mid-2026, the newest of the first-party Wan task models), and its stock ComfyUI workflow is built from a grab-bag of text-manipulation nodes - the author's comment in the source calls out exactly what this replaces: a CustomCombo + RegexExtract + StringReplace + StringConcatenate chain. This node collapses that whole text-wrangling group into two dropdowns.

How it works

Two inputs, two outputs:

• dance_style - 古典舞 (classical), 韩舞 (K-pop), 街舞 (street), 拉丁舞 (Latin), 踢踏舞 (tap), defaulting to classical.
• motion_amplitude - 低 (low), 中等 (medium), 高 (high), 最大 (maximum), defaulting to medium.

The first output, 舞蹈风格提示词 ("dance style prompt"), is just the style name - it's meant to replace the RegexExtract output that the original workflow fed into a prompt node. The second output, 整体描述提示词 ("overall description prompt"), is the assembled full prompt: something like 一个人正在跳舞，舞蹈种类是古典舞,图像清晰程度高,人物动作幅度中等 - "a person is dancing, the dance type is classical, image clarity is high, the motion amplitude is medium."

Both strings are Chinese, because the original Wan Dancer workflow and its prompts are Chinese. If your pipeline prompts in English, either feed these into a translation node or be ready to hand-edit - the switcher is faithful to the stock workflow, not anglicized.

Why you'd use it

If you're building a Wan Dancer project - a character dancing to a music track - this node turns "change the dance" from "rewire the text pipeline" into "change one dropdown." It also makes batch exploration trivial: switch style, queue, compare, repeat. The outputs are plain STRINGs, so they wire into any text-encode node.

Install & notes

Identical to the rest of the pack: ComfyUI Manager → XB_ToolBox, or git clone https://github.com/wjluoxiao/XB_ToolBox.git into custom_nodes, restart. No models, no dependencies beyond the pack's declared opencv-python and easyocr.

Two honest notes. First, the dance variety lives in the model and your audio, not this node - the switcher only sets two words of the prompt. Second, Wan-Dancer itself is brand new as of this writing with essentially no community reception yet, so treat the whole workflow as "the author's workflow," not "community-proven." This node is the reliable, boring part of that.