Black Forest Labs' LoRAs are taken from
FLUX.1-Canny-dev-lora and
FLUX.1-Depth-dev-lora, and pruned using
resize_lora.py in Kohya's sd-scripts. They're in a different format from SDXL ControlLoRAs, so they should be loaded in
LoraLoaderModelOnly node and used in
InstructPixToPixConditioning node, see the
example workflow.
Sometimes Shakker Labs' union ControlNet can follow the control better than Black Forest Labs' models, but it's in another different format and I need more time to figure out how to prune and load it.
