How do I speed up Qwen Image generation?

Lower the image resolution or reduce the number of steps in the KSampler settings. Using fp8 model files (already recommended here) also significantly reduces VRAM usage and generation time.

How to Use Qwen Image in ComfyUI (2026 Step-by-Step Guide)

Q: Is Qwen Image free to use locally?

Yes. The model weights are available for free on Hugging Face and can be run locally in ComfyUI without any subscription or API cost.

Qwen Image is a powerful text-to-image model that turns words into stunning images. With this guide, it's easy for anyone to install Qwen Image locally and use it with ComfyUI. Follow these simple steps to get started and see your ideas come to life!

What is Qwen Image?

Quick answer: Qwen Image is an advanced text-to-image AI model by the Qwen team. It generates detailed, high-quality images from text prompts and runs fully locally in ComfyUI — no subscription or API cost required.

Qwen Image is an advanced text-to-image AI model created by the Qwen team. It can generate detailed and creative images from simple text prompts. Qwen Image is popular for its ability to create high-quality pictures, making it useful for artists, designers, and anyone who wants to visualize ideas from text.

The model runs as a local ComfyUI workflow using three separate model files: a diffusion model, a text encoder (Qwen 2.5 VL 7B), and a VAE. All three are available for free download from Hugging Face.

Download Qwen Image Workflow

To use Qwen Image in ComfyUI, download the ready-made workflow file below. This workflow is pre-configured with all the correct node connections — just load it and select your model files.

💡 Tip: Download and load this workflow first before placing model files. It will show you exactly which nodes need which files.

Download Qwen Image Workflow

Required Models

Before running the workflow, download these three Qwen Image model files and place them in the correct folders of your ComfyUI installation.

Model Files

Diffusion Model: qwen_image_fp8_e4m3fn.safetensors
Text Encoder: qwen_2.5_vl_7b_fp8_scaled.safetensors
VAE: qwen_image_vae.safetensors

Folder Structure — Where to Place Model Files

Make sure your folders look like this:

ComfyUI/
└─ models/
   ├─ diffusion_models/
   │   └─ qwen_image_fp8_e4m3fn.safetensors
   ├─ vae/
   │   └─ qwen_image_vae.safetensors
   └─ text_encoders/
       └─ qwen_2.5_vl_7b_fp8_scaled.safetensors

⚠️ Warning: Using the wrong folder for a model file is the most common cause of errors. Double-check each file is in the exact folder shown above before running the workflow.

How to Install Qwen Image Locally (Step-by-Step)

Follow these eight steps to get Qwen Image running inside ComfyUI on your local machine.

Update ComfyUI

Always use the latest version of ComfyUI for best compatibility with new models. Pull the latest commit or re-download from the official GitHub repository.

Download the Workflow

Get the workflow JSON file from the download button in the section above.

Download the Three Model Files

Download the diffusion model, text encoder, and VAE files from the links above. Place each file in its correct folder as shown in the folder structure.

Load the Workflow in ComfyUI

Open ComfyUI in your browser. Drag and drop the downloaded workflow JSON file into the ComfyUI canvas to load it.

Select Model Files in the Workflow Nodes

In the Load Diffusion Model node, select qwen_image_fp8_e4m3fn.safetensors. In the Load CLIP node, select qwen_2.5_vl_7b_fp8_scaled.safetensors. In the Load VAE node, select qwen_image_vae.safetensors.

Enter Your Prompt

Type your image description in the workflow's prompt box. See the example prompts in the next section for inspiration.

Adjust Settings (Optional)

Change image width, height, or advanced settings like CFG, steps, and shift values for different results. Default settings work well for a first test.

Click "Run" and Wait

Click the Queue Prompt / Run button. Wait for Qwen Image to generate your AI image. Generation time depends on your GPU.

Example Run

Here are three real examples generated with Qwen Image in ComfyUI. Each card shows the exact prompt used and the output image — click any image to view it full size.

Example 1 — Fashion Portrait — Prompt Used

A high-fashion editorial shot of a realistic young woman with flawless makeup and styled hair. The image emphasizes ultra-sharp textures such as fabric weave on her avant-garde dress, the sparkle in her eyes, and the slight sheen on her lips. The lighting is studio quality, highlighting every detail crisply with a slight vignette around the edges. The background is minimalist white to keep all focus on the subject.

Qwen Image ComfyUI output — high-fashion editorial portrait with ultra-sharp textures and studio lighting — Output result — click to view full size

Example 2 — Dark Fantasy — Prompt Used

A dark, hooded figure stands in the midst of thick fog. The creature's eyes glow with hellish fire, as if the darkness itself is pouring through them. Its face is covered in shadow, but from beneath the dark purple hood, a vicious yellow skull can be seen, twisted with malice. A clawed hand reaches forward, as if calling or cursing. Magical energy swirls around it, like smoke, poisoning the air with an ominous aura.

Qwen Image ComfyUI output — dark hooded figure with glowing eyes and swirling magical energy in thick fog — Output result — click to view full size

Example 3 — Outdoor Scene — Prompt Used

At the lush boundary where a dense, humid jungle meets a sun-bleached beach, a beautiful woman seeks refuge from the heat. She leans back against the rough bark of a palm tree, her skin dewy with sweat. The dappled sunlight filtering through the canopy above creates a complex pattern of light and shadow across her glistening skin. The camera reveals the vibrant, steamy jungle environment behind her, emphasizing the raw, untamed heat of the moment.

Qwen Image ComfyUI output — woman at jungle and beach boundary with dappled sunlight and tropical atmosphere — Output result — click to view full size

Tips & Troubleshooting

If something is not working as expected, one of these is almost always the cause:

Image Not Generating — model files not found or in wrong folder

Fix: Double-check that all three model files are in the exact correct folders (diffusion_models, text_encoders, vae) and that each is selected in the correct node inside the workflow.

Slow Generation Performance

Fix: Lower the image resolution or reduce the number of steps in the KSampler. The fp8 model files are already optimized for lower VRAM — make sure you downloaded the fp8 variants.

Results Not as Expected

Fix: Try more descriptive prompts. Adjust CFG (guidance scale) — higher values follow the prompt more closely. Also try adjusting the shift value for different compositional results.

Model Files Not Appearing in Node Dropdowns

Fix: After placing new model files in the folders, press F5 or Ctrl+Shift+R to fully refresh ComfyUI in your browser. Without this step, the new files will not appear in dropdowns.

💡 Tip: For a wider list of ComfyUI errors and fixes, see the ComfyUI Troubleshooting Guide.

Final Thoughts

Qwen Image makes it simple to turn text descriptions into high-quality images with just a few steps. Install it locally, follow this guide, and explore new creative possibilities with this advanced text-to-image model. The fp8 model files keep VRAM usage manageable even on mid-range GPUs, making Qwen Image accessible to a wide range of hardware setups.

Frequently Asked Questions

Qwen Image is an advanced text-to-image AI model created by the Qwen team. It generates detailed, high-quality images from text prompts and runs fully locally in ComfyUI without any subscription or API cost.

Place qwen_image_fp8_e4m3fn.safetensors in ComfyUI/models/diffusion_models, qwen_2.5_vl_7b_fp8_scaled.safetensors in ComfyUI/models/text_encoders, and qwen_image_vae.safetensors in ComfyUI/models/vae.

Check that all three model files are in the correct folders and selected in the correct workflow nodes. Also verify ComfyUI is updated to the latest version and refresh your browser after adding new files.

Lower the image resolution or reduce the number of steps in the KSampler settings. The fp8 model files (already recommended in this guide) also significantly reduce VRAM usage and generation time.

Yes. All model weights are available for free on Hugging Face and can be run locally in ComfyUI without any subscription, API key, or ongoing cost.