Unlock the New Qwen-Image GGUF Workflow for Low-VRAM Editing! | CGPIXEL AI on Patreon
Unlock the New Qwen-Image GGUF Workflow for Low-VRAM Editing!
CGPIXEL AI
Aug 19, 2025
Hey everyone, I’m thrilled to share a powerful, streamlined Qwen-Image workflow I’ve just released—perfect for users with low-VRAM systems and designed to make image editing with simple prompts easier than ever.
What’s New?
GGUF-formatted Qwen-Image model for low-VRAM setups
Modular Node-Based Workflow in ComfyUI Using ComfyUI (now offering native support for Qwen-Image models as of August 6, 2025).
Simple Prompt-Driven Editing The workflow supports tasks like style transfers, object insertion/removal, precise text editing—and even multilingual text rendering (e.g., Chinese, English) within images.
Choose Your Variant Wisely
- From ComfyUI Wiki benchmarks: the fp8 variant uses ~86% of an RTX 4090’s 24 GB VRAM and generates an image in ~71 seconds (after the first run).
- If you have less VRAM, try using smaller quant variants like Q5 or Q4 as described.
Guide Recap:
- Download the GGUF model (e.g., qwen-image-Q8_0.gguf), text encoder (e.g., qwen_2.5_vl_7b_fp8_scaled.safetensors), and VAE (qwen_image_vae.safetensors).
- Place them in the respective folders in your ComfyUI installation (models/diffusion_models, text_encoders, vae).
- Update ComfyUI (use nightly or latest version to ensure full Qwen-Image support)
- Load the nodes: Model, Text Encoder, VAE—then plug in prompt nodes and run!
- Use simple prompts to invoke editing features—from text insertion to style shifts to object manipulations.
Ready to Try?
Grab the workflow files, model variants. Let’s bring creative control back to our hands—even on resource-limited hardware. More exclusive content for paid subscribers including (prompting, upscaling and fine tuning results)
Happy editing!
Attachments
Download qwen img editing gguf (1).json