【QWEN】INPAINT — v2.0

【QWEN】INPAINT is a ComfyUI workflow for Qwen inpainting and text-to-image generation. It takes a Qwen model, written prompt, and optional LoRA settings, then produces images with generation parameters embedded as metadata.

The auto version supports automatic model downloads, and the built-in SeedVR2 tiled upscaler can be toggled on or off.

Three LoRA slots offer individual on/off toggles and strength controls; additional LoRA modules can be connected.

Use a Qwen model and written prompt; inpainting and LoRA controls are also supported.

The workflow outputs images with all generation parameters embedded as metadata.

Enable ComfyUI’s Nodes 2.0 (Vue) interface before loading; otherwise display problems may occur.

Install the specified custom node, ComfyUI-UmeAiRT-Toolkit, through ComfyUI Manager by searching “UmeAiRT,” or use the UmeAiRT Auto-Installer.

Required upscaler model: seedvr2_ema_7b_sharp_fp8_e4m3fn_mixed_block35_fp16.safetensors. Its availability is not confirmed.

Required node pack: comfyui-umeairt-toolkit. Its availability is not confirmed.

For manual setup, use fp8 or bf16 model variants in models\diffusion_models, or Q8, Q5, or Q4 GGUF variants in models\unet.

For the base version, place qwen_2.5_vl_7b_fp8_scaled.safetensors in models\clip; GGUF versions list CLIP options Q8, Q6, Q5, Q4, and Q3.

Common files are Qwen-Image-InstantX-ControlNet-Inpainting.safetensors in models\controlnet, qwen_image_vae.safetensors in models\vae, and either Qwen-Image-Lightning-8steps-V2.0.safetensors or Qwen-Image-Lightning-4steps-V2.0.safetensors in models\loras.

Before enabling SeedVR2, verify that seedvr2_ema_7b_sharp_fp8_e4m3fn_mixed_block35_fp16.safetensors is available.

Vorschlag · nicht geprüft

Before loading the workflow, verify that comfyui-umeairt-toolkit is available.

Vorschlag · nicht geprüft