Flux 2 Klein 9B Edit Character FaceSwap Image-to-Image, Low VRAM(6 GB)

The Flux Klein 9B Edit Model by Black Forest Labs is one of the most efficient and visually striking image-to-image diffusion models currently available for creative AI workflows. Designed for high-quality edits with natural tones and detailed facial reconstruction, it brings cinematic realism to every image. When combined with the BFS Character Face Swap Model by Alissonerdx/BFS-Best-Face-Swap, you can easily perform seamless, high-fidelity character swaps that retain lighting, expression, and texture accuracy — all within ComfyUI.
 
This setup goes a step further by integrating the Flux Klein 9B Edit model with specialized LoRA modules such as bfs_head_v1_flux-klein_9b and ultra_real_v3 to enhance facial adaptation and realism. Using these LoRAs allows smoother blending between faces while retaining art style or realism control sliders for precision results. The combination produces AI-generated portraits, cosplay swaps, or cinematic photoreal imagery ideal for creators working on storytelling or content design projects.
 
The default workflows provided by the creators are already optimized for high VRAM systems. However, my custom workflow has been redesigned specifically for low VRAM machines by leveraging the GGUF versions of both the Klein model and the CLIP text encoder. This optimization ensures you can run image-to-image or face-swap operations locally — even on mid-range GPUs — without visibly losing quality.
Default Workflows - https://huggingface.co/Alissonerdx/BFS-Best-Face-Swap/tree/main/workflows
 
With the One-Click Flux Klein 9B Installer, everything comes preconfigured for immediate use, saving hours of manual setup.
What you’ll get with the installer
A one-click Windows installer that sets up ComfyUI, installs all dependencies, and preloads models automatically. Just launch, wait for setup, and start creating images instantly.
A low-VRAM optimized workflow using GGUF models that still deliver stunning, photorealistic renders.
 
Preloaded Models (Low VRAM-Ready):
flux-2-klein-9b-Q5_K_M.gguf – UNet model
Hugging Face Link
Qwen3-8B-UD-Q5_K_XL.gguf – CLIP text encoder
Hugging Face Link
flux2_vae.safetensors – VAE (diffusion decoder)
Hugging Face Link
2xLexicaRRDBNet_Sharp.pth – Upscaler
Hugging Face Link
bfs_head_v1_flux-klein_9b_step3500_rank128.safetensors – Face swap LoRA
Hugging Face Link
ultra_real_v3.safetensors – LoRA for enhanced realism
Hugging Face Link
klein-9b_realism_slider_v2.safetensors – Realism control slider
Hugging Face Link
 
Note on diffusion models
The standard Flux 2 Klein 9B BF16 diffusion checkpoints are not bundled in the installer to keep the package lightweight, but they can be downloaded from the official Black Forest Labs collection:
Flux 2 Klein 9B Diffusion Models
 
​​​​​​​Custom Nodes and Commands for Portable Setup
ComfyUI Manager: https://github.com/ltdrdata/ComfyUI-Manager
Install requirements with:
.\python_embeded\python.exe -m pip install -r .\ComfyUI\custom_nodes\ComfyUI-Manager\requirements.txt
Use the ComfyUI Manager’s “Install missing custom nodes” option if any nodes fail to load after installation.
 
System requirements
Nvidia RTX 30XX/40XX/50XX GPUs (FP16 support recommended)
CUDA-compatible GPU with 6GB+ VRAM (8GB+ ideal)
Windows OS (64-bit)
At least 30GB of free storage
What’s included
Fully portable ComfyUI Windows installer pre-configured for Flux 2 Klein 9B Edit character face swapping
Automated model and node downloads
User-friendly workflows suitable for beginners or pros
Fast Groups Bypasser support for one-click workflow switching
 
​​​​​​​Usage notes
Load the provided workflow into ComfyUI.
Verify all model checkpoints are mapped correctly in your Loader nodes (especially your Character LoRA).
Upload a reference face and a target image/body through the “Load Image” node for seamless AI face-swapping.
Keep resolutions under 1024×1024 to prevent VRAM spikes on lower-end GPUs.