A ComfyUI node inserted between the model/Lora loader and the KSampler that changes how Flux renders an image, independent of seed and prompt. Exposes four controls: Max Shift and Base Shift and Width/Height. Placement relative to the Lora doesn't matter — before or after both work the same.
Gives finer control over Flux outputs: lets you fix specific problem details or vary composition without a full re-roll (i.e. without changing the seed). Especially useful paired with Flux Turbo Lora (8-step workflow), where it helps claw back detail lost to the compressed step count.
No single ideal configuration exists; tuning is empirical (see ComfyUI Batch Parameter-Sweep Workflow for a systematic way to explore it).
Place the node in the workflow chain between the checkpoint/Lora loader and the KSampler, then tune Max Shift, Base Shift, Width, and Height before sampling to fix or vary details while keeping seed and prompt fixed.