Multi-Model Workflow - Qwen 2511 + Wan 2.2 — v1.0

This workflow creates images from text prompts or transforms existing images, with an optional reference image; Qwen ImageEdit 2511 handles composition and Wan 2.2 T2V adds detail to the resulting image.

Qwen ImageEdit 2511 replaces Wan's high-noise composition stage, while Wan 2.2's low-noise stage adds detail.

For text-to-image, provide a text prompt; for image-to-image, start with an existing image, and you can enable a reference image in either mode.

Use an existing image to add detail or carry a character or scene composition into Wan's detail stage.

Use one image to generate multiple-style iterations; LoRAs can make this easier.

Download it from https://civitai.com/models/2449354; it is listed as a Workflows item by DirkDigitaller with Qwen as the base model.

Listed LoRA files include ComicRealism01a_CE_QWEN_AIT3k.safetensors, Duality02_CE_QWEN_AIT3k.safetensors, facetest-000007.safetensors, ghibli_style_qwen_v3.safetensors, low_noise_model.safetensors, pencil-sketch-anime-qwen.safetensors, and SmoothMixStyle_Low.safetensors.

Other listed model files include wan_2.1_vae.safetensors, wan_masterpieces_v2.safetensors, qwen_masterpieces_v3.safetensors, qwen_2.5_vl_7b_fp8_scaled.safetensors, qwen_image_edit_2511_fp8_e4m3fn_scaled_lightning_comfyui_4steps_v1.0.safetensors, qwen_image_vae.safetensors, umt5_x

Named node packs include comfyui-easy-use, comfyui-kjnodes, comfyui-rmbg, RES4LYF, rgthree-comfy, and was-ns.

Use the options panel to enable or disable Reference Image, and use Qwen/Wan Specific Prompting to target only the relevant model with certain details.

Vorschlag · nicht geprüft

Use Wan NAG to tell Wan what to focus on, and Reactor FaceSwap to reinforce Wan's character consistency.

Vorschlag · nicht geprüft

Use the workflow notes to adjust settings toward increased consistency or increased divergence.

Vorschlag · nicht geprüft

If you use SageAttention, turn it off globally and toggle Patch SageAttention in the workflow because Qwen does not work well with the global setting.

Vorschlag · nicht geprüft

Check the qwen_image_vae.safetensors filename against your setup; if you use a different Qwen ImageEdit base model, consider downloading the 4-step LoRA to keep steps and speed consistent.

Vorschlag · nicht geprüft