NitroFusion2-Controlnet-Union

NitroFusion2-Controlnet-Union is a finetuned version of Z-Image-Turbo-Fun-Controlnet-Union-2.1 (2602), adapted to NitroFusion2 for 1-step and 2-step generation. It must be used with NitroFusion2 as the base model.

Supported conditions: Canny, HED, Depth, Pose, Scribble, Gray, and inpainting.

ComfyUI

1 step: workflows/nitrofusion2_controlnet_1step.json
2 steps: workflows/nitrofusion2_controlnet_2step.json

The 2-step workflow loads the LoRA with ComfyUI-BypassLoRA. It applies the LoRA without merging it into the weights, which makes 2-step generation 1.5–3.5× faster. The node includes the fix from ComfyUI PR #17408 until ComfyUI ships it.

Model Card

Name	Description
NitroFusion2-Controlnet-Union.safetensors	Union 2602 (15 control layer blocks + 2 refiner blocks) adapted to NitroFusion2 for 1–2 step inference.
NitroFusion2-Controlnet-Union-lite.safetensors	Lite version (3 control layer blocks + 2 refiner blocks) adapted the same way; smaller and lighter on memory, with weaker control.

Results

1 step

Pose	Union 2.1-2602 + NitroFusion2	NitroFusion2-Controlnet-Union

Pose	Union 2.1-2602 + NitroFusion2	NitroFusion2-Controlnet-Union

Canny	Union 2.1-2602 + NitroFusion2	NitroFusion2-Controlnet-Union

Canny	Union 2.1-2602 + NitroFusion2	NitroFusion2-Controlnet-Union

HED	Union 2.1-2602 + NitroFusion2	NitroFusion2-Controlnet-Union

HED	Union 2.1-2602 + NitroFusion2	NitroFusion2-Controlnet-Union

Depth	Union 2.1-2602 + NitroFusion2	NitroFusion2-Controlnet-Union

Depth	Union 2.1-2602 + NitroFusion2	NitroFusion2-Controlnet-Union

Pose + Inpaint	Union 2.1-2602 + NitroFusion2	NitroFusion2-Controlnet-Union

2 steps

Pose	Union 2.1-2602 + NitroFusion2	NitroFusion2-Controlnet-Union

Pose	Union 2.1-2602 + NitroFusion2	NitroFusion2-Controlnet-Union

Canny	Union 2.1-2602 + NitroFusion2	NitroFusion2-Controlnet-Union

Canny	Union 2.1-2602 + NitroFusion2	NitroFusion2-Controlnet-Union

HED	Union 2.1-2602 + NitroFusion2	NitroFusion2-Controlnet-Union

HED	Union 2.1-2602 + NitroFusion2	NitroFusion2-Controlnet-Union

Depth	Union 2.1-2602 + NitroFusion2	NitroFusion2-Controlnet-Union

Depth	Union 2.1-2602 + NitroFusion2	NitroFusion2-Controlnet-Union

Pose + Inpaint	Union 2.1-2602 + NitroFusion2	NitroFusion2-Controlnet-Union

Citation

If you find NitroFusion2 useful in your research, please cite our paper:

@article{chen2026nitrofusion2,
    title={NitroFusion2: Bringing Adversarial Loss to the Text Stream of Multimodal Diffusion Transformers},
    author={Chen, Dar-Yen and Song, Yi-Zhe},
    journal={arXiv preprint arXiv:2611.05648},
    year={2026}
}