Action Modifier (Single)

Action Modifier is the dimension that decides whether your subject is frozen, moving, racing, or parked. Twenty descriptions covering motion, scenario, and state - in motion, arriving, departing, stationary display, suspended motion, being driven, racing, cruising - each appended to your base prompt by ID. It's the pack's answer to the difference between a car sitting in a lot and a car tearing down a road, and that distinction is worth a lot more in an image than people usually give it credit for.

The modifiers read like a cinematographer's shot list: 1 is "in motion, dynamic movement, kinetic energy captured," 4 is "stationary display, static presentation, parked positioning," 5 is "suspended motion, frozen moment, peak action captured," 7 is "racing/competing, high speed action, competitive scene," 8 is "cruising leisurely, relaxed motion, casual movement." The top of the range covers interactions and presentations. Because the vocabulary is concrete and physical, it's one of the dimensions where a single description can visibly change the body language of the subject.

The inputs: text (base prompt), modifier_id (1–20, default 7), and preserve_base (default true). The default of 7 (racing) is a strong default - fine for the automotive shots the pack clearly was built around, but double-check it if your renders keep coming out mid-race when you wanted a parked display. Output: the usual single STRING, modified_prompt.

Install is the pack-standard zero-dependency story: ComfyUI Manager → search "Variationator" → install → restart, or cd ComfyUI/custom_nodes && git clone https://github.com/MushroomFleet/ComfyUI-Variationator.git. No models, no requirements.txt to install (the repo's install-portable.bat references one that doesn't ship). Node appears under Prompt Orchestration.

Where this node pays off is in combination with camera and lighting: "racing" + "low angle" + "golden hour" is the classic hero shot, and the pack's whole architecture is designed for exactly that kind of stacking - every node outputs STRING so you can chain them in sequence. Two honest notes. First, action words conflict easily: appending "stationary display" to a base that already says "driving through the city" produces a contradiction the encoder has to resolve, so reach for preserve_base=False when your base already implies motion. Second, keep action to one clause per prompt - "in motion, racing, kinetic energy, dynamic movement, high speed action" is four ways to say the same thing, and that's how you burn attention budget on redundancy. One action, well-chosen, beats five.