Qwen-Image 2.1 BFS Character Swap

Requirements

• NVIDIA RTX 30XX / 40XX / 50XX GPU
• CUDA-compatible GPU — ~8 GB for int8, or 6-8 GB with the GGUF turbo quant
• At least 20 GB free storage

Model downloads

• Alissonerdx https://huggingface.co/Alissonerdx/BFS-Best-Face-Swap ⧉
• bfs_head_v1.1_alternative_qwen_2.1.safetensors — in the loader, switched off (stronger expression transfer, slightly less accurate head angle) https://huggingface.co/Alissonerdx/BFS-Best-Face-Swap/tree/main ⧉
• qwen_image_2.1_turbo_int8_convrot.safetensors – the default https://huggingface.co/Comfy-Org/Qwen-Image-2.1/tree/main/diffusion_models ⧉
• qwen_image_2.1_turbo_Q5_K_M.gguf – low VRAM: turn on "Enable GGUF Model", off "Enable Diffusion Model" https://huggingface.co/Abiray/Qwen-Image-2.1-Turbo-GGUF/tree/main ⧉
• qwen3vl_8b_int8_convrot.safetensors https://huggingface.co/Comfy-Org/Qwen-Image-2.1/tree/main/text_encoders ⧉
• qwen_image_2.1_vae_bf16.safetensors – required https://huggingface.co/Comfy-Org/Qwen-Image-2.1/tree/main/vae ⧉
• Qwen/Qwen-Image-2.1-Turbo https://huggingface.co/Qwen/Qwen-Image-2.1-Turbo ⧉

Custom nodes

• ComfyUI Manager https://github.com/ltdrdata/ComfyUI-Manager ⧉
• ComfyUI-GGUF (GGUF loader path) https://github.com/city96/ComfyUI-GGUF ⧉
• rgthree-comfy (Power Lora Loader, bypasser) https://github.com/rgthree/rgthree-comfy ⧉
• RES4LYF (for the bong_tangent scheduler) https://github.com/ClownsharkBatwing/RES4LYF ⧉
• ComfyUI-KJNodes (side-by-side preview) https://github.com/kijai/ComfyUI-KJNodes ⧉

Setup steps

1. Download the free workflow .json with the button above.
2. Download the 7 model file(s) listed above into their ComfyUI model folders.
3. Install the 5 custom node(s) listed above via ComfyUI Manager, then fully restart ComfyUI.
4. Drag the workflow .json onto the ComfyUI canvas and confirm each loader node has the right model selected.
5. Run it. If a node shows red, update ComfyUI and restart before troubleshooting anything else.

Notes & tips

• The BFS LoRAs (Qwen 2.1 head versions): bfs_head_v1.1_qwen_2.1.safetensors — loaded by default, recommended by the author (less skin smoothing, best head-angle fidelity) — Hugging Face Repo; bfs_head_v1.1_alternative_qwen_2.1.safetensors — in the loader, switched off (stronger expression transfer, slightly less accurate head angle) — Hugging Face Repo
• Settings: Steps 8, CFG 1, sampler lcm, scheduler bong_tangent (bong_tangent comes with the RES4LYF node pack).; CFG 1.2 was more than twice as slow in testing with no gain in resemblance — stay at 1.; LoRA strength 1.0.
• Use a sharp, front-facing head reference.
• Use a body image where the face is clearly visible (waist-up or closer) — faces far away in full-body shots swap poorly.
• Author's caveat: Qwen 2.1 can shift the whole image slightly (colours, softness) even for small edits; higher resolutions help.
• The built-in side-by-side preview stacks your body, head and result so you can check the swap at a glance.
• 🔒 Members Exclusive + Video — Now Live: The exclusive enhanced workflow; A one-click installer that sets it all up for you; A full walkthrough tutorial video
• More Qwen 2.1: Qwen-Image 2.1 All-in-One One-Click Installer (Patreon) - https://www.patreon.com/TheLocalLab/posts/qwen-image-2-1-170300735

More resources

• https://www.patreon.com/TheLocalLab/posts/qwen-image-2-1-172314062 ⧉
• Qwen-Image 2.1 All-in-One One-Click Installer (Patreon) https://www.patreon.com/TheLocalLab/posts/qwen-image-2-1-170300735 ⧉