How should I handle negative conditioning with Qwen?

For Qwen image editing, it is recommended to always use the ConditioningZeroOut node for negative conditioning rather than standard text-based negative prompts.

How do I achieve the fastest generation speeds?

Use the 4-step Lightning LoRA with a CFG of 1.0, Euler sampler, and simple scheduler to generate high-quality edits in just four steps.

Can I use Qwen Image Edit for video resolutions?

Yes, it supports specific resolutions like 832x480, which are natively compatible with video generation pipelines like WAN 2.2 FLF.

Which models are required for Qwen Image Edit workflows?

The workflow requires three primary components: the Qwen 2511 edit UNET (bf16), the shared Qwen 2.5 VL CLIP (fp8), and the specific Qwen VAE.

What is the advantage of the TextEncodeQwenImageEditPlusAdvance node?

The Advanced node is preferred because it outputs a latent directly, eliminating the need for a separate EmptyLatentImage or VAEEncode node, and provides superior control over image resizing.

Qwen Image Edit Workflows

Name: Qwen Image Edit Workflows
Author: artokun

byartokun

•

データサイエンスとML

Constructs and optimizes sophisticated Qwen2.5-VL image editing pipelines in ComfyUI using natural language instructions.

This skill empowers Claude to build, configure, and troubleshoot advanced image editing workflows based on the Qwen2.5-VL vision-language model. It provides deep technical guidance on model selection—including the official 2511 edit model and community realism remixes—advanced conditioning nodes for direct latent output, and Lightning LoRA integration for high-speed generation. Users can leverage this skill to implement precise image transformations, multi-angle consistency patterns, and automated XY plot testing for optimal results.

主な機能

01XY plot implementation for batch-testing prompt variations and model parameters

02Multi-angle instruction patterns with SKS tokens for consistent view generation

03Lightning LoRA integration for high-quality 4-step and 8-step image generation

0411 GitHub stars

05Automated setup for Qwen2.5-VL 2511 edit and hyper-realistic community models

06Advanced conditioning with direct latent output and automatic image resizing

ユースケース

01Transforming specific image elements or subjects via natural language instructions

02Building multi-view asset sheets for character design and 3D modeling pipelines

03Generating hyper-realistic product photography and professional portrait edits

What are Skills?·How to Install

Install with 🐟 Skill.Fish

npx skillfish add artokun/comfyui-mcp qwen-image-edit

For use in Claude.ai and ChatGPT

Download Skill