QWEN Image Edit - Ultimate 2509 edit

Use QWEN Image Edit - Ultimate 2509 edit to edit a named subject or region in an image: provide an image and a prompt naming what to segment, and it can create an adjustable mask and cropped region for sampling.

Load an image and write a prompt naming what to segment, such as a character, girl, or car.

Adjust the segmented region, crop and resize it for sampling, and use the reference latent or light LoRA controls.

Input: an image and a prompt naming the subject to segment, such as a character, girl, or car.

When box mask is enabled, the workflow provides a box around the segmented character, and the sampler works on the cropped region.

Begin in Init by loading the image and writing the prompt.

Suggestion · non vérifié

The listed model files are: 4x-AnimeSharp.pth, qwen_2.5_vl_7b_fp8_scaled.safetensors, qwen_image_edit_2509_fp8_e4m3fn.safetensors, qwen_image_vae.safetensors, Qwen-Image-Lightning-4steps-V1.0.safetensors, and sam2_hiera_large.pt.

Named node packs include comfyui_essentials, comfyui_layerstyle, ComfyUI-Allor, comfyui-crystools, comfyui-custom-scripts, comfyui-detail-daemon, comfyui-easy-use, comfyui-inpaint-cropandstitch, and comfyui-kjnodes.

Additional named node packs include ComfyUI-QualityOfLifeSuit_Omar92, comfyui-sam2, efficiency-nodes-comfyui, masquerade-nodes-comfyui, rgthree-comfy, and was-ns.

If you enable box mask, adjust the box's X and Y position, width, and height; set the cropped region's total megapixels, or set the Resize node to False.

Set manual mask growth or blur growth, adjust the segmentation threshold, and choose whether to use the light LoRA or add more LoRAs.

Before use, verify that these listed items are available: Qwen-Image-Lightning-4steps-V1.0.safetensors, 4x-AnimeSharp.pth, sam2_hiera_large.pt, and was-ns.

Suggestion · non vérifié

Treat the reference latent as experimental: its results can vary with the model, light LoRA, and crop, and can sometimes be worse.

Suggestion · non vérifié