☁️BizyAir Apply Style Model

This is where a Redux-style image gets blended into your conditioning. You've loaded a style model, you've run your reference image through CLIP Vision, and now this node takes both of those plus your existing text conditioning and merges them into one conditioning output that carries both the words and the visual style.

The community's honest read on Redux-style conditioning, worth knowing before you build a workflow around it: it's strong. Strong enough that people building custom workflows around it found the default setup produced results that were mostly variations of the input rather than something the prompt could meaningfully steer. If you've tried this style of node and felt like your text prompt got completely steamrolled, that's a known, reported behavior, not something you're doing wrong - it just means you're looking for finer control over how much the style model contributes, which is exactly what BizyAir Apply style model (simple) exists to give you as a friendlier alternative to raw strength values. This node, by contrast, is the advanced version: no strength knob at all, just a clean merge.

Inputs and outputs

All three inputs are required:

• conditioning - your existing text conditioning, type BIZYAIR_CONDITIONING.
• style_model - the loaded style model, type BIZYAIR_STYLE_MODEL, which has to come from BizyAir Load Style Model upstream.
• clip_vision_output - the encoded reference image, type CLIP_VISION_OUTPUT. This is the one input in this node that uses a standard ComfyUI type rather than a BizyAir-specific one, so it should connect fine from a normal CLIP Vision Encode node.

The output is BIZYAIR_CONDITIONING - a single merged conditioning that carries both your prompt and the style model's influence. Feed it into your sampler the same way you would any conditioning, in place of (not alongside) your plain text conditioning.

Installing BizyAir

Through ComfyUI Manager, search "BizyAir" and install. Or by hand:

cd ComfyUI/custom_nodes && git clone https://github.com/siliconflow/BizyAir.git

Restart ComfyUI, or run comfy node install bizyair on Comfy-CLI. Set your API key on first use via the README's "click to login" link - without it, the merge this node performs simply won't run, since it's executed on BizyAir's cloud rather than locally.

Common issues

If your output looks like a near-copy of the reference image and ignores your prompt, that's the well-documented Redux behavior rather than a bug - the fix is either to lower the strength somewhere upstream (if your style model or workflow exposes one) or switch to the "simple" variant of this node, which trades a raw numeric strength for named presets (highest down to lowest) that are easier to reason about.

Watch your types here too: conditioning and style_model both need to be BIZYAIR_CONDITIONING/BIZYAIR_STYLE_MODEL, meaning they have to originate from other BizyAir nodes. Only clip_vision_output accepts a standard ComfyUI type. If a connection refuses to make in the graph editor, that's almost always a BizyAir-vs-stock type mismatch, not a broken node.