Speedy Wan 2.2 Controlnet Using Flux (Nunchaku Compatible) — v1.0

Use this workflow to generate initial images in Flux with a ControlNet you choose, then refine those images with WAN 2.2.

Choose a ControlNet for Flux; the described results are initial Flux images and their WAN 2.2 refinements.

You’ll need the LoRA [lightx2v_T2V_14B_cfg_step_distill_v2_lora_rank256_bf16.safetensors](https://huggingface.co/Kijai/WanVideo_comfy/resolve/main/Lightx2v/lightx2v_T2V_14B_cfg_step_distill_v2_lora_rank256_bf16.safetensors?download=true).

Named node packs include comfyui-kjnodes, comfyui-impact-pack, ComfyUI-GGUF, rgthree-comfy, comfyui-nunchaku, and RES4LYF.

Named model files include Wan2.2-T2V-A14B-LowNoise-Q4_K_M.gguf, t5xxl_fp8_e4m3fn.safetensors, umt5_xxl_fp8_e4m3fn_scaled.safetensors, wan_2.1_vae.safetensors, and lightx2v_T2V_14B_cfg_step_distill_v2_lora_rank256_bf16.safetensors.

Other named files include diffusion_pytorch_model.safetensors, flux.1 turbo alpha.safetensors, flux1DevVAE_safetensors.safetensors, and Long-ViT-L-14-BEST-GmP-smooth-ft.safetensors.

For the initial image stage, choose which ControlNet Flux should use.

Suggestion · not verified

Before downloading, confirm which ControlNet file is intended by diffusion_pytorch_model.safetensors.

Suggestion · not verified

Before installing, verify which ComfyUI-GGUF package is intended.

Suggestion · not verified