Home Creators Posts Import Register Favorites Logout

Downloads

Content

This is really the best image editing model you can download and run on your machine right now.

Below you can find the manual instructions for installing. As a supporter you can download my Easy installer that will handle the model download, custom node installation and workflow setup for you.⬇️ https://www.patreon.com/AIKnowledgeCentral/posts/qwen-image-2-1-170561864

🐍 Python & PyTorch

This workload was built and tested with:

Python 3.12

PyTorch 2.11.0 with CUDA 13.0

📦 ComfyUI & custom nodes

Start with ComfyUI itself:https://github.com/comfyanonymous/ComfyUI

Then install these custom nodes into ComfyUI\custom_nodes (or search for them by name in ComfyUI-Manager):

ComfyUI-IntoTheLatent-Utils

comfyui_controlnet_aux

ComfyUI-Manager

rgthree-comfy

cg-use-everywhere

🎨 Models

🖼️ Diffusion models → ComfyUI\models\diffusion_models

qwen_image_2.1 – pick the one that fits your graphics card's VRAM:

8 GB: qwen_image_2.1_int8_convrot.safetensors

24 GB: qwen_image_2.1_bf16.safetensors

📝 Text encoders → ComfyUI\models\text_encoders

qwen3vl_8b – pick the one that fits your graphics card's VRAM:

8 GB: qwen3vl_8b_int8_convrot.safetensors

24 GB: qwen3vl_8b_bf16.safetensors

🎭 VAE → ComfyUI\models\vae

qwen_image_2.1_vae_bf16

qwen_image_2.1_vae_bf16.safetensors

⬇️ Download the workflows below this post ⬇️

Files

Forget Cloud Editors: Use Qwen 2.1 Instead!

Qwen 2.1 can take up to ten input images at once. It removes backgrounds, upscales blurry pictures and, with a little trick, even works with ControlNet, all for free on your own PC. After testing it, this is now my go-to model for image editing. In this video I walk you through three ComfyUI workflows step by step. We start with text to image and test how well Qwen 2.1 handles photorealism, recognising characters, real RGBA transparency and long text without typos. Then we move to the image edit workflow, where the model really shines. I show you how to use my multi-image loader and the custom resolution switch. Then we change colours while keeping every detail, re-pose a character, cut subjects out onto a transparent background and combine five reference images into one consistent character. You'll also learn how to upscale a low-res image properly, and why you have to describe the missing details in your prompt. I explain what LoRAs are, where to put them and how to use trigger words and strength. Finally, I show you how to use an OpenPose ControlNet to copy the pose and camera angle from a reference image. I also explain how to switch on the Nodes 2.0 beta, which you need for the image compare node. All workflows are a free download on my Patreon. 📩 Stay up to date with Local AI & Open Source AI https://intothelatent.com/newsletter ❤️ Support the channel & get exclusive workflows, guides and early access https://patreon.com/intothelatent" 📅 Book a personal 2-hour 1-on-1 AI coaching session (paid offer) where we'll solve your specific workflow, ComfyUI, Stable Diffusion, or local AI questions together. https://koalendar.com/e/1hr-1-on-1-stable-diffusion 🏠 My Website https://intothelatent.com Chapters 00:00 Intro & Setup 01:02 Text 2 Image - Testing Photorealism, Characters & Real Transparency 05:43 Long Text Without Typos 06:35 Precise Image Edits, New Poses & Background Removal 11:04 Multi Input Images into One Scene 13:40 Upscaling Blurry Images, LoRAs & ControlNet 15:44 Using Loras 18:30 ControlNet Results & Final Verdict

Comments

No comments found for this post.