How to Build FLUX Multi-Reference Character Workflow

Roshpatoich · reddit · 2026-07-15

The author wants to build a "3-in-1" character generation workflow for FLUX in ComfyUI. The goal isn't a face swap, but native integration during the diffusion process:

Text-only mode: Standard text-to-image.
Text + Face Reference: Uses a single face image to preserve identity features, with text dictating clothing, pose, and scene.
Text + Face + Body Reference: Uses a high-quality face image as the identity and style source, and a full-body shot as a "body blueprint," extracting only body type, silhouette, clothing shape, and pose without copying its original style.

They originally used ComfyUIPuLIDFluxll for identity preservation, but updating to the latest ComfyUI Portable caused multiple API compatibility issues, including transformeroptions, attnmask, and timestepzeroindex. The author wants to confirm:

Is PuLID Flux LL still the best node for native identity preservation?
Is anyone maintaining it for the latest ComfyUI?
Would Flux Redux, IPAdapter FaceID, PuLID, or other combinations be better?

The core requirement: one image defines "who," another defines "how the body poses," text defines actions and environment, and FLUX natively generates a unified, natural character image.

Original post →

How to Build FLUX Multi-Reference Character Workflow

More from Multimodal