Wan 2.2 Detailer

Wan 2.2 Detailer is a workflow for adding detail to an input image with the Wan 2.2 T2V low-noise model. Enter an image and composition prompt to create a detailed image, with options for focus and character consistency.

Load your image into the LoadImage node and describe its composition in the Prompt text box; an LLM can help write the prompt.

Suggestion · non vérifié

The workflow includes directions for refining the result toward greater consistency or greater divergence from the original image.

The main inputs are an image and a composition prompt, and the output is a more detailed version of the image.

The workflow is published by DirkDigitaller and uses Wan Video 2.2 T2V-A14B as its base model.

Model files named by the workflow are facetest-000007.safetensors, low_noise_model.safetensors, SmoothMixStyle_Low.safetensors, umt5_xxl_fp8_e4m3fn_scaled.safetensors, wan_2.1_vae.safetensors, wan_masterpieces_v2.safetensors, and wan2.2_t2v_low_noise_14B_fp8_scaled.safetensors.

Node packs named by the workflow are comfyui_layerstyle, comfyui-kjnodes, comfyui-rmbg, RES4LYF, rgthree-comfy, and was-ns.

Use Wan NAG to direct focus toward the image elements you want emphasized.

Suggestion · non vérifié

Use Reactor FaceSwap when you want increased character consistency from the original image.

Suggestion · non vérifié

Before using the workflow, verify that the named model files and node packs are available in your setup.

Suggestion · non vérifié