This is a very simple ComfyUI beginner friendly QWEN Image Edit Plus (2509) Simple Head Swap GGUF Workflow that will work with very simple text editing instructions in natural language to swap head of your desired target image's subject from a reference image. You need a Hugging Face account to download your necessary QWEN Image files (Details are mentioned below) for this workflow. Make sure you install GGUF addon for ComfyUI using ComfyUI manager and place the correct files in correct places. Also check out my other workflows for SD 1.5 + SDXL 1.0, HiDream, QWAN, Chroma, WAN, KREA and Flux.
How to use this -
#1. Just select your desired QWEN Image model first
#2. now select your image(s) on "Upload Image(s) for Editing" - a reference face image & a target image for face swap
#3. on next step enter your image editing instructions
#4. then select how many images you want (Change the number besides the "Run" button)
#5. after this select image sampling methods, CFG, steps etc. settings (you may wanna stay with default
#5. finally press the run button to generate. That's it..
***Update for Version 2.0 -
This version no longer gives you bad faces on good head swap, this will also do perfect face swaps (with notes on how to perform NSFW swaps too on "How to Enable Batch NSFW Faceswap" text file included with the workflow) with some extra steps. For Face Swap you can either also select the same reference face image that you used for head swap or select a different face (this is why I added an extra Image upload option)
Recommended Settings for this Workflow
===========================================================
Sampler - euler_a
Scheduler - Beta57
CFG - 1.4 (You can change between 1.2 to 1.5)
Lora Strength - 1.4 (You can change between 1.2 to 1.5)
Steps - 14 (You can change between 12 to 15)
Just a reminder adding Qwen Lightning LoRA (4 or 8 Steps) will likely to give your result's subject "plastic"-ish skin,
adding other LORAs may alter the results in unexpected ways (good or bad).
## QWEN Image Edit Models
=============================================================
### Download Links for QWEN Image Edit Checkpoint
### Download Links for QWEN Image Encoders
### Download Links for QWEN Image VAE
https://huggingface.co/QuantStack/Qwen-Image-Edit-GGUF/resolve/main/VAE/Qwen_Image-VAE.safetensors
==============================================================
### This workflow also uses the following QWEN Image LORA from CivitAI -
BFS - Best Face Swap - https://civarchive.com/models/2027766?modelVersionId=2304596
Description
***Update for Version 2.0 -
This version no longer gives you bad faces on good head swap, this will also do perfect face swaps (with notes on how to perform NSFW swaps too on "How to Enable Batch NSFW Faceswap" text file included with the workflow) with some extra steps. For Face Swap you can either also select the same reference face image that you used for head swap or select a different face (this is why I added an extra Image upload option)
FAQ
Comments (8)
***Update for Version 2.0 -
This version no longer gives you bad faces on good head swap, this will also do perfect face swaps (with notes on how to perform NSFW swaps too on "How to Enable Batch NSFW Faceswap" text file included with the workflow) with some extra steps. For Face Swap you can either also select the same reference face image that you used for head swap or select a different face (this is why I added an extra Image upload option)
Will this work on low vram?
Yes it is designed to work on 8 GB VRAM.. I can't say if it will work on 6 GB VRAM or lower.. you should at least have 8 GB VRAM (this is a must)
Hello, will this work with Qwen 2511?
Yes, you just have to swap the files to their updated matching QWEN 2511 versions
@sarcastictofu Thanks for your answer, i see i need KJnodes and Inpaint CropAndStitch can you tell me wich version i need?
Also, the execution is very slow on my RTX 5070 ti, is it normal?
@estarossa9441 I can not tell anything about RTX because I use all AMD system (Ryzen AI 7 Processor with AMD Radeon GPU) running Linux but I do have newer Flux.2 Klein versions ( one for 4B and one for 9B ) of this workflow that you can find on my profile.. these are significantly faster and in many situations (not always) offer better results.. 4B one you can even run on with 4 GB to 6 GB VRAM as some folks have told me.. (my own GPU has 8 GB VRAM)