H3 Creative Presets

H3 Creative Presets is the quiet utility of the colorAi MiniMax H3 pack, and it does exactly one thing: it builds a reusable "creative brief" object that the pack's prompt nodes fold into their LLM request. No API calls, no GPU, no model weights - it's pure configuration. You could live without it, but if you generate the same kind of thing repeatedly - brand promos, MV subtitles, claymation shorts - this is how you stop retyping "3D CG style, neon night lighting, low-key cinematic" into every prompt.

How it works

Every dropdown on the node corresponds to a field in the H3 prompt's creative contract: visual_style, environment, time_weather, lighting, shot_size, camera_angle, camera_motion, motion_amplitude, motion_speed, shot_structure, soundscape, and music, plus a free-text custom_instructions box for lens, continuity, or sound rules the menus don't cover. Pick your values and the node packs them into a single H3_CREATIVE_PRESET that wires into the creative_presets input on either the MiniMax H3 Prompt Studio + Generate node or the legacy Prompt Engineer · RunningHub node.

The neat bit: leave any dropdown on Auto and it's simply omitted - the LLM infers that value from your user_request and reference material instead. So you can lock down the style you care about and let the request decide the rest. A preset isn't a hard override of your prompt; it's an injected layer of direction the model writes against.

The one setting that matters

Honestly, most of the node is friction-free defaults. The two things worth changing from the start: visual_style if you have a consistent look, and custom_instructions - that's where your real creative constraints live, since it's the only free-text field. Everything else earns its keep when you need consistency across a series of clips: a "brand promo" preset means every run in that batch carries the same lighting and camera language without you re-arguing with the LLM.

Installing and wiring it

It ships in the same pack as the prompt nodes, so install once and you have all three: search "MiniMax H3 Prompt Engineer" in ComfyUI Manager, or git clone https://github.com/colorAi/ComfyUI-Minimax-H3-Prompt-Engineer into custom_nodes/, then restart. The only dependency is certifi. In your workflow it's an input provider, not an output node - build the preset, feed it to the prompt node's creative_presets socket, and reuse the same preset node in as many workflows as you like. There's little to troubleshoot because there's little to fail: no network, no models, no keys. It's the rare ComfyUI node you can trust to just work.