☁️BizyAir ControlNetLoader_Legacy

"Legacy" is doing real work in this node's name. Where the current BizyAir_ControlNetLoader gives you a live catalog dropdown plus a model_version_id field for pulling in custom or newly-added models, this one has neither - just a fixed, hardcoded list of specific ControlNet files. It's kept around for workflows built against those exact files before the newer loader existed; for anything new, reach for the non-legacy loader instead, unless you specifically need one of the models only listed here.

What's actually on the list is worth knowing, because it's a genuinely useful cross-section of the ControlNet landscape rather than an arbitrary sample. sdxl/diffusion_pytorch_model_promax.safetensors is xinsir's SDXL union ControlNet under its "promax" filename - the single most capable ControlNet artifact most people will ever load, covering twelve conditions (canny, depth, pose, segmentation, normals, and more) in one file, still wider than anything trained for a newer base as of 2026. sd15/control_v11f1e_sd15_tile.pth is the SD 1.5 tile model, the foundation of classic tiled-upscaling workflows. instantid/diffusion_pytorch_model.safetensors is InstantID's IdentityNet - a ControlNet-shaped module built specifically for identity preservation rather than general structure conditioning. kolors/Kolors-ControlNet-Canny.safetensors and Kolors-ControlNet-Depth.safetensors cover Kwai's Kolors base model. And the three sd3.5_large_controlnet_* entries (blur, canny, depth) are Stability's own official ControlNets for SD3.5 Large.

The practical takeaway: this list spans several genuinely different base architectures under one dropdown, so - same as the current loader - the thing that matters most is picking the entry that actually matches the checkpoint you're generating with. A Kolors ControlNet does nothing meaningful paired with an SD3.5 checkpoint; the SDXL union only works with an SDXL-family base. There's no cross-compatibility to lean on here regardless of how good any individual entry is.

Inputs that matter: control_net_name, required, a fixed enum - the eight entries listed above (as of this pack's current release; BizyAir may extend the legacy list over time, but don't expect it to track new architectures the way the primary loader does). No other configurable fields - no model_version_id, no device switch, nothing.

Output: CONTROL_NET in the display label, BIZYAIR_CONTROL_NET under the hood - same as the primary loader, only compatible with BizyAir's ControlNet apply nodes downstream.

Install: search "BizyAir" in ComfyUI Manager and install, git clone https://github.com/siliconflow/BizyAir.git into custom_nodes and restart, or comfy node install bizyair - one pack install covers both ControlNet loaders and everything else BizyAir ships. It's a cloud call like the rest of the pack, so a registered BizyAir account with your API key set is required before this loader will do anything, dropdown included.

Troubleshooting: if you're reaching for this node instead of the primary loader, first double-check you actually need something specific to this fixed list - if what you want is available through the regular BizyAir_ControlNetLoader's live catalog, that one's the better default going forward. Beyond that, the same rules apply as everywhere in this pack: architecture mismatches between the loaded ControlNet and your base checkpoint are the most common source of "no visible effect," and any stall or unclear failure is worth an API-key check before you go looking elsewhere.