Qwen Image Edit Plus

Qwen Image Edit Plus is the upgraded editor in the pack's Qwen lineup, and its selling point is a model you've probably heard of in a different context: Qwen Edit 2511. That's the late-2025 open-weight editor that the KB panel credits with becoming the base for most character-consistency tooling - and this node's model dropdown pairs it with "Qwen Edit Plus" itself, with the tooltip calling out multi-person editing and improved consistency as the 2511 model's strengths.

If you've ever edited a group photo with an instruction model, you know exactly why that matters. Most editors handle a single subject well and then melt faces, swap identities, or flatten two similar-looking people into one when multiple characters are in frame. The 2511 line was the one that got meaningfully better at keeping several distinct people consistent across an edit. That's the niche this node owns.

The inputs

• model - Qwen Edit Plus or Qwen Edit 2511
• prompt - the modification, in Chinese or English
• images - up to 3 reference images, comma-separated URLs or paths
• width / height - 256 to 1536 (step 8)
• output_format - output image format (default jpeg)
• enable_sync_mode / enable_base64_output - completion and encoding behavior
• seed - reproducibility, -1 for random
• client - optional WaveSpeed client

The output is an image IMAGE tensor. Structure is identical to the Qwen Image 2.0 Edit node - the meaningful difference is the model you're calling, not the wiring. For multi-character work, pick 2511; for general edits, either serves.

Install and setup

cd ComfyUI/custom_nodes
git clone https://github.com/eRepublik-Labs/comfyui-nodes-erpk.git erpk
cd erpk && pip install -r requirements.txt

Or "ERPK Custom Nodes" from ComfyUI Manager, restart, and put a WaveSpeed API key in ERPK Settings. The WaveSpeed key is the only credential this needs - the model runs hosted, billed per image.

Troubleshooting and where it fits

Same family gotchas as the other Qwen edit nodes: images is a comma-separated string (not a tensor), the cap is 3, and it's a whole-frame editor - unmasked pixels will drift, and drift compounds across a chain of edits. If you're doing a multi-step character edit, that KB advice applies harder than ever: do one change at a time, and for anything surgical, mask around it rather than trusting a single long prompt.

Worth saying plainly: if your only edit is "add a hat to this one guy in a crowd," the 2511 model is the one you want - that's precisely the multi-person consistency case it was built for. If you're not sure whether you need it, run the same edit on the standard Qwen Edit model first; the difference shows up exactly when it matters.