FLUX-Kontext Portrait Grid Emulation in ComfyUI (Dynamic Prompts + Switches for Low RAM) — v1.0

Allows model/clip/VAE switching for low vs normal RAM setups using Any Switch (rgthree)

Includes pose preservation and face consistency across all outputs

Batch text injection + seed control

Optional face swap and background removal tools included

Que up 12 and make sure the text number is at zero (see screen shots) it will cycle through the prompts. You of course can make better prompts if you wish. The image makes a black background but you can change that to whatever color you wish.

lastly there is a faceswap to improve on the end results. You can delete it if you are not into that.

This is all thanks you BGRated who requested such a workflow. who did this for me on Matteo's Discord server. Thank you zGenMedia you rock.

📦 Node Packs Used:

rgthree-comfy (for switches & group toggles)

comfyui_layerstyle (for dynamic text & image blending) NOTE: Download this model (model.pth) and place it ComfyUI/custom_nodes/ComfyUI_LayerStyle/RMBG-1.4/

comfyui-itools (for pose string rotation)

comfyui-multigpu (for Flux-Kontext compatibility)

comfy-core (standard utilities)

ReActorFaceSwap (optional FaceSwap block)

ComfyUI_LayerStyle_Advance (for PersonMaskUltra V2)

⚠️ Heads Up:

This isn’t the most elegant setup—prompt logic can still be refined, and pose diversity may need manual tweaks. But it’s usable out the box and should give you a working foundation to tweak further.

📁 Download & Screenshots:

[ Workflow: https://pastebin.com/v8aN8MJd ] Just remove the txt at the end of the file if you download it.

Grid sample and pose output previews attached below are stitched by me the program does not stitch the final results together.

Any help or support can be done here . It literally puts food on my table. If you need Runpod credits any new joins get $500 for $10 install.

Checkpoint · 7.71 GB · GGUF · Unknown