☁️BizyAir MinusZone - KolorsControlNetLoader

The name says Kolors, and that's honestly where this node started, per the README it's part of BizyAir's August 2024 Kolors integration, ported from MinusZoneAI's ComfyUI-Kolors-MZ. But look at what's actually in the control_net_name dropdown and it's grown well past a single-architecture loader: alongside kolors/Kolors-ControlNet-Canny.safetensors and kolors/Kolors-ControlNet-Depth.safetensors, the catalog includes SDXL checkpoints (sdxl/diffusion_pytorch_model_promax.safetensors), SD 1.5 (sd15/control_v11f1e_sd15_tile.pth), InstantID (instantid/diffusion_pytorch_model.safetensors), and the full SD 3.5 Large ControlNet trio for blur, canny and depth. It's the general-purpose cloud ControlNet checkpoint loader for this pack, wearing a name that reflects its origin more than its current scope.

What this node does, and doesn't, do

This is purely a loader, it doesn't run any preprocessing and it doesn't apply the ControlNet to your model, it just resolves control_net_name to a hosted checkpoint and hands back a BIZYAIR_CONTROL_NET you wire into a downstream apply/conditioning node elsewhere in a BizyAir workflow. The actual conditioning image still needs to come from a preprocessor, one of this pack's own depth, lineart, or M-LSD nodes, or another BizyAir preprocessor upstream in your graph.

Matching the checkpoint to your base model

The path-style naming in the dropdown is the whole guide here: the prefix tells you exactly which base architecture that ControlNet checkpoint targets, kolors/ for a Kolors model, sdxl/ for SDXL, sd15/ for SD 1.5, instantid/ for an InstantID identity-conditioning setup, and the bare sd3.5_large_controlnet_* entries for SD 3.5 Large. ControlNet checkpoints are architecture-bound the same way LoRAs are, loading a Kolors ControlNet against an SDXL model (or vice versa) won't produce a usable result, it isn't a compatibility layer, just a resolver. Make sure whatever BIZYAIR_MODEL you've loaded elsewhere in the graph actually matches the prefix on the checkpoint you pick here.

Worth knowing if you're coming to this fresh: this catalog reflects an earlier generation of ControlNet coverage, InstantID and SD 3.5's canny/depth/blur trio, alongside SDXL and SD 1.5 staples. If you're working on a newer base architecture outside this list, you'll want that architecture's own dedicated ControlNet union rather than looking for it here.

Installing it

Through ComfyUI Manager: search "BizyAir", install, restart. Or clone directly:

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

You'll need a BizyAir/SiliconFlow API key set up before this node resolves anything, via the "click to login" prompt the README points to on first use. Signup involves phone-number verification, and it's worth knowing ahead of time that some users outside China have reported friction there.

Common issues

If the ControlNet appears to load but has no visible effect on the generation, the most common cause is an architecture mismatch between the checkpoint prefix and the base model you loaded, double-check kolors/, sdxl/, sd15/, instantid/ or sd3.5_large_* actually lines up with what's feeding your sampler. A conditioning image that doesn't match what the checkpoint expects (a depth map fed into a canny-trained checkpoint, for instance) will also produce weak or wrong results without an outright error.

As with every node in this pack, a run that stalls or fails with no clear workflow-level cause almost always traces back to the API key, check that first. Separately: if BizyAir shows up in your custom_nodes folder without you having installed it, that's a known ComfyUI Manager quirk (misresolving an unrelated missing-node error on an older shared workflow), not malware, and it's safe to remove through the Manager.