Nodes/ComfyUI_agilly1989_motorway/ControlNetApplySD3_motorway_edition
ComfyUI Node

ControlNetApplySD3_motorway_edition

ControlNet for SD3/Flux-style models, wired through the Motorway

By agilly1989·Created 2 years ago·Updated about a year ago· 7
ControlNetApplySD3_motorway_edition
  • MOTORWAY 🚌💨
  • MOTORWAY 🚌💨
INPUT_positive_keypositive
INPUT_negative_keynegative
INPUT_control_net_keycontrol_net
INPUT_vae_keyvae
INPUT_image_keyimage
strength1.00
start_percent0.000
end_percent1.000
OUTPUT_positive_keypositive
OUTPUT_negative_keynegative

ControlNetApplySD3 is the ControlNet apply node for the MMDiT generation - SD3, Flux, and their descendants - the ones whose ControlNets expect the conditioning image encoded by the model's own VAE before they'll do anything sensible. The _motorway_edition wraps that core node so all of its inputs ride in the Motorway bus by key instead of on wires.

Why it exists separately

With SD1.5/SDXL ControlNets, the image encoding happens inside the apply node against a VAE you pass in. The SD3/Flux-era apply node is the same idea but tuned for how those models were trained: the reference ControlNets for SD3 and Flux all consume a VAE-encoded image, and this is the node that carries the VAE input to make that happen. In practice it's why Flux ControlNet workflows look different from SDXL ones - the loaders and apply nodes are different classes even though the concept is identical. If you're running an SD3 or Flux base, this (or the Advanced variant of it) is the apply node you want; the classic ControlNetApplyAdvanced targets the UNet-era architectures.

Inputs and outputs

  • MOTORWAY 🚌💨 - the bus, required.
  • INPUT_positive_key / INPUT_negative_key - your two conditionings, defaults positive / negative.
  • INPUT_control_net_key - the loaded CONTROL_NET, default control_net.
  • INPUT_vae_key - the model's VAE, used to encode the condition image, default vae.
  • INPUT_image_key - the condition image, default image.
  • strength - control weight, 0–10, default 1.0.
  • start_percent / end_percent - the apply window, 0–1.
  • OUTPUT_positive_key / OUTPUT_negative_key - where results land, defaults positive / negative.

The output is the Motorway; pull the conditioned pair back out with a Motorway AxB ramp node.

Installing

The node ships in the whole pack:

cd ComfyUI/custom_nodes
git clone https://github.com/agilly1989/ComfyUI_agilly1989_motorway

Restart, or install via ComfyUI Manager ("ComfyUI_agilly1989_motorway"). No pip deps. You'll need an SD3/Flux-compatible ControlNet file and the matching VAE in your models folder.

Caveats

Pack-wide reality: active beta, "BIG BROKEN WITH ASYNC" README banner, and the _motorway_edition clones are commented out of __init__.py in the version I checked - re-enable the ClonedNodeMapping import if the node's missing from the menu. Keys are exact and case-sensitive (hash_ error = key not in bus). One honest take: Flux ControlNets are thinner on the ground than the SD-era menu - the KB's controlnet essay is blunt that the modern bases got canny/depth/pose and lost the exotics - so before wiring a whole bus around this node, confirm the ControlNet you want actually exists for your base. If you're not on a Motorway graph, plain ControlNetApplySD3 is the same node with less setup.

Categoryagilly1989 Nodes/Motorway-ed/conditioning/controlnet

Inputs (11)

NameTypeDefaultDescription
MOTORWAY 🚌💨MOTORWAY 🚌💨
INPUT_positive_keySTRINGpositive
INPUT_negative_keySTRINGnegative
INPUT_control_net_keySTRINGcontrol_net
INPUT_vae_keySTRINGvae
INPUT_image_keySTRINGimage
strengthFLOAT1.000–10
start_percentFLOAT0.0000–1
end_percentFLOAT1.0000–1
OUTPUT_positive_keySTRINGpositive
OUTPUT_negative_keySTRINGnegative

Outputs (1)

NameTypeDescription
MOTORWAY 🚌💨MOTORWAY 🚌💨