What is the fastest way to generate images with Qwen 2512?

Using the 4-step Lightning LoRA with the integrated sampler allows for high-quality generation in approximately 3-5 seconds per image.

Which sampler settings are best for realistic portraits?

For ultra-realistic portraits, use the CopaxTimeless variant with 30 steps, a CFG of 4.0, and the res_multistep sampler with sgm_uniform scheduler.

Does this skill support ControlNet for Qwen models?

Yes, it supports QwenImageDiffsynthControlnet for Canny, Depth, and Inpaint tasks, allowing you to patch models with specific control signals.

What resolution should I use for Qwen Image 2512?

Qwen operates natively at roughly 1.6 megapixels; recommended sizes include 1328x1328 for square, 1056x1584 for 2:3 portraits, and 1664x928 for 16:9 widescreen.

How do I handle negative prompts with Qwen?

Qwen performs best with natural language prompts. For negative conditioning, use the ConditioningZeroOut node or provide an empty string for the negative input.

Qwen Image 2512 Workflows

Name: Qwen Image 2512 Workflows
Author: artokun

byartokun

•

データサイエンスとML

Generates high-fidelity text-to-image workflows for Qwen 2512 models, supporting lightning-fast sampling and advanced component loading.

This skill empowers Claude to architect and execute sophisticated image generation pipelines using the Qwen Image 2512 model family within ComfyUI. It provides domain-specific knowledge for both the streamlined QwenImageIntegratedKSampler and modular component loading, enabling precise control over UNET, CLIP (Qwen2.5-VL), and VAE configurations. Users can leverage detailed guidance on 4-step and 8-step Lightning LoRAs for near-instant results, apply AuraFlow sampling shifts for standard quality, and implement specialized fine-tuned models for hyper-realism and artistic styles. It also includes comprehensive support for native resolutions up to 1.6 megapixels and DiffSynth-based ControlNet integration.

主な機能

01Automated setup for Integrated KSampler and modular component pipelines

02Advanced sampling controls including AuraFlow shift and zero-out conditioning

03Built-in support for fine-tuned models like UltimateRealism and CopaxTimeless

04Pre-configured 4-step and 8-step Lightning LoRA acceleration

05Optimized resolution presets for 1.6MP native Qwen output

0611 GitHub stars

ユースケース

01Rapidly generating high-quality marketing visuals using natural language prompts

02Building complex image-to-image and ControlNet pipelines for precise character posing

03Prototyping realistic product photography with specialized fine-tuned model variants

What are Skills?·How to Install

Install with 🐟 Skill.Fish

npx skillfish add artokun/comfyui-mcp qwen-txt2img

For use in Claude.ai and ChatGPT

Download Skill