☁️BizyAir Apply style model (simple)

Same job as BizyAir Apply Style Model - merge a Redux-style visual conditioning into your existing text conditioning - but with one extra input that makes it noticeably friendlier for a first attempt: image_strength, a plain-English dropdown instead of a raw numeric weight.

That's a real usability fix, not just a cosmetic one. The core problem with Redux-style conditioning is that it can dominate the prompt almost entirely - early community testing on FLUX.1 Redux found the useful strength range was narrow, something like 0.04 to 0.07 on a 0–1 scale before the model stopped listening to text at all and just gave you variations of the reference image. Remembering "0.06 is about right" is much harder than picking "low" from a list, which is exactly the gap this node closes.

How it works

It does the same conditioning merge as the non-simple version, but instead of a numeric strength value, image_strength snaps to one of five preset levels. Behind the scenes this presumably maps to sensible numeric values already tuned for this kind of workflow - you just pick where on the spectrum from "barely there" to "dominant" you want the style model's pull to sit.

Inputs and outputs

• conditioning - your text conditioning, type BIZYAIR_CONDITIONING.
• style_model - the loaded style model, type BIZYAIR_STYLE_MODEL, from BizyAir Load Style Model.
• clip_vision_output - your encoded reference image, type CLIP_VISION_OUTPUT, the one standard-typed input here (a normal CLIP Vision Encode node feeds this fine).
• image_strength - an enum with five choices: highest, high, medium (the default), low, lowest. Start at medium and move toward low/lowest if the style is overpowering your prompt, or toward high/highest if you want the reference image's look to dominate more strongly.

Output is BIZYAIR_CONDITIONING, ready to feed straight into your sampler in place of your plain text conditioning.

Installing BizyAir

Through ComfyUI Manager: search "BizyAir," install. Or clone it yourself:

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

Restart ComfyUI, or use comfy node install bizyair on Comfy-CLI. Set your API key first - the README's "click to login" step is required before this or any other BizyAir node will actually run, since the merge happens on SiliconFlow's cloud.

Common issues

If your result still ignores your text prompt even at low or lowest, that tells you the issue is likely elsewhere in the workflow - check that you're not also feeding a very high-weight style model or stacking multiple style-conditioning passes, since those compound.

Going the other direction, if highest still isn't giving you enough of the reference image's character, that's a sign you may want the non-simple Apply Style Model node instead, in case your use case needs finer control than five presets provide - though for most first attempts, this node's default of medium is the more forgiving starting point of the two.

As always, the type constraints matter: conditioning and style_model both require BizyAir-typed sources, so this whole chain needs to stay inside the pack rather than mixing in stock ComfyUI conditioning nodes partway through.