☁️BizyAir LeReS Depth Map (enable boost for leres++)

LeReS is one of the older depth estimators in the ControlNet preprocessor lineup, from around the same era as MiDaS, and it's a fine choice for depth ControlNet if you don't feel like fighting with a Depth Anything install. This node is BizyAir's cloud version of the same LeReS preprocessor you'd find in the standard comfyui_controlnet_aux pack, same parameters, same behavior, just run remotely on SiliconFlow's servers so you never download the model weights yourself.

What it does

Feed it an image and it hands back a grayscale IMAGE, brighter pixels closer to camera, darker further away, the standard depth-map convention every ControlNet depth model expects. That map goes straight into a depth ControlNet loader and conditioning node from there; this preprocessor's whole job is extraction, not conditioning.

The knobs worth knowing

rm_nearest and rm_background are the two settings that make LeReS distinctive versus a plain MiDaS pass. Turn them up and the preprocessor strips out the closest and farthest depth values respectively before returning the map, useful when your image has a distracting near-camera object or a background you'd rather the ControlNet ignore entirely. Leave them at 0 (the default) for a normal full-scene depth pass.

boost, when enabled, switches on what's informally called "leres++", a higher-quality mode that trades speed for a sharper, more detailed map. It's off by default; turn it on when the plain pass looks too soft, particularly on complex scenes with lots of fine geometry. resolution is the usual preprocessing-resolution tradeoff, 512 is a reasonable default, push it toward 1024 if you need the depth map to hold up on a large final render, at the cost of processing time.

Where it fits

Wire the output IMAGE into whatever depth ControlNet loader and apply-node your base workflow uses, this preprocessor doesn't care which architecture the ControlNet is trained for, it just produces a standard depth map. Typical ControlNet weight for a depth condition runs 0.4-0.7 for loose guidance that leaves room for the model to invent detail, or 0.8-1.1 when you need the spatial layout to hold tightly, and ending the ControlNet's influence around 40-60% through the steps rather than running it the whole way tends to give cleaner results either way.

Installing it

Through ComfyUI Manager: search "BizyAir" and install, then restart. Or via git:

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

Because this node runs on SiliconFlow's cloud rather than locally, you'll need a BizyAir API key set up before it does anything, the README points you to a "click to login" prompt on first use. Signup goes through phone-number verification via SiliconFlow.

Common issues

If the run doesn't execute at all, check the API key before anything else, that's the single most common failure across every node in this pack. If the depth map comes back looking flat or missing detail on a busy scene, try boost before assuming the model itself is the problem, leres++ genuinely produces a noticeably sharper result on complex geometry. And if a subject that should read as "close" is coming through washed out or missing, check whether rm_nearest or rm_background got left at a nonzero value from a previous workflow, those two settings are easy to forget you touched.

Separately from anything this node does: ComfyUI Manager has been known to auto-install the entire BizyAir pack when it misresolves an unrelated missing-node error on an older shared workflow. That's a real, if confusing, quirk of the Manager's node-registry inference rather than anything malicious, and it's safe to remove if you didn't intend to install it.