This ComfyUI workflow is a powerful DeepFace / Face + Body Reference tool built on Flux.2 Klein (9B or 4B).

It uses two reference images:

• Body Reference (full body pose)

• Face Reference (close-up face)

The workflow combines them using Reference Conditioning to generate consistent characters with accurate face swapping, body pose preservation, and high-quality photorealistic results.

Perfect for:

• Consistent character generation

• Face swapping / head swapping

• Outfit or pose control

• Creating training data for LoRAs

Main Features:

• Dual reference system (Face + Body)

• Strong face and pose adherence

• Background removal option (RMBG)

• Image resizing and conditioning

• Clean output with Flux.2 Klein speed and quality

Required Models:

• Flux.2 Klein (recommended: Flux2Klein_9B or 4B base / fp8)

• Flux VAE

• CLIP (Qwen-2.5 or similar for Flux)

• Optional LoRA support

Created by: bobgus39 Original profile: https://civarchive.com/user/bobgus39

Usage Tips: Load your Body Reference and Face Reference images, adjust the prompt if needed, and run. The system will respect the original pose, camera angle, lighting, and location while transferring the face accurately.

Great for character consistency and dataset creation!