LLPS Controller: workflow-level declaration for LLPS preview management.
LLPS Manager panel: frontend workflow scanner for sampler-like nodes and LLPS-controlled status.
LLPS Config (Legacy) and LLPS KSampler (Legacy): deprecated v1.2 compatibility nodes for callback-based preview saving.

v2-alpha design boundary

This version does not globally override every sampler in the workflow yet. The LLPS Controller is the workflow-level authority used by the Manager panel to classify sampler-like nodes as controlled, uncontrolled, or candidate.

LLPS KSampler remains available as a legacy/prototype node for v1.2 callback-based preview saving, but it is not the architectural center of LLPS v2.

Install

Copy this folder into:

ComfyUI/custom_nodes/ComfyUI-LLPS-v1/

Restart ComfyUI.

Basic use

Add LLPS Controller.
Open the LLPS Manager panel.
Use Refresh to scan sampler-like nodes in the workflow.
Use the panel to inspect which sampler-like nodes are covered by the enabled Controller.

For legacy v1.2 preview saving, LLPS Config can still be connected to LLPS KSampler.

LLPS Manager panel

After restarting ComfyUI, use the floating LLPS button or the LLPS menu to open the manager panel.

The panel scans the current workflow and shows:

LLPS nodes
sampler-like nodes such as KSampler, KSamplerAdvanced, SamplerCustom, SamplerCustomAdvanced, and Ultimate SD Upscale nodes
whether each sampler is LLPS-controlled, uncontrolled, or a candidate
Controller coverage status for sampler-like nodes
node id, node type, node title
status filters, Focus, Refresh, and uncontrolled-node selection actions

When the panel is open, matching nodes are visually marked on the canvas.

Settings

live_preview_method: server_default, none, auto, latent2rgb, taesd
show_preview: send live preview frames to ComfyUI UI
save_preview: save the same preview frames to disk
save_base_path: empty = ComfyUI/output/LLPS
subfolder: optional explicit subfolder
filename_prefix: prefix for step files
image_format: jpg, png, webp
save_every_n_steps: save interval; 1 saves every step
save_metadata_json: writes metadata.json next to saved frames

Important

none means no preview frame is decoded, so it cannot save frames. For “save only,” use:

live_preview_method = latent2rgb or taesd
show_preview = false
save_preview = true