MiniMax System Prompt Preset

Prompt presets for video-generation models are a weird corner of this ecosystem: the system prompt is often half the battle, and the good ones are long, mode-specific strings that nobody wants to retype. MiniMax System Prompt Preset packages those strings for MiniMax's prompting modes into a single node - pick a mode, get the system prompt, wire it into a Generate node's system input.

The type combo is the selector, defaulting to I2V (image-to-video). The modes are I2V, FL2V, FL2V_LOOP, T2V, R2V, R2I, R2A, and L2V - that's your image-to-video, frame-latent-to-video (and its loop variant), text-to-video, and the reference modes (reference-to-video/image/audio) plus audio-related L2V. The node description explains the composition rule: reference modes use their reference base prompt, other modes use the common base, and the selected base and mode prompt are joined with one blank line.

The mechanism is deliberately simple - it loads a packaged prompt string. What's interesting is the override path: the optional enum_string input, when connected, overrides type and must be one of the exact mode names. That's the string-based pathway - handy if you want to drive the mode from elsewhere in the graph (a text node, a saved workflow, an automated batch) rather than the combo widget. The tooltip spells out the allowed values, so there's no guessing about spelling.

The output is a single system prompt string, with a tooltip that tells you exactly what to do with it: connect it to a Generate node's system input. If you're using this pack's Ollama Generate or llama.cpp Generate, that's the natural target - you get the model-appropriate MiniMax guidance without pasting a paragraph from a README.

One caution that applies to any preset node: these strings are tuned for MiniMax's prompting conventions, and the prompt is sent verbatim - the node won't translate your I2V mode choice into a generic "describe what changes" instruction for a different model. Use it when you're actually driving MiniMax, not as a universal prompt source.

Install is the pack standard:

cd ComfyUI/custom_nodes
git clone https://github.com/craftingmod/ComfyUI-Ollama-ImageList

Restart ComfyUI. No extra dependencies beyond the pack itself and ComfyUI 0.19.3+ for the V3 API.

A genuinely thin node with a genuinely useful payoff: it turns "memorize which MiniMax prompt format goes with which mode" into a dropdown. If you iterate between T2V and I2V setups, the reference-vs-common-base distinction is handled for you, and enum_string gives the automation crowd a clean string hook. Not glamorous - but the day you'd rather not hand-write a system prompt for a mode you use once a month, it's the node you'll reach for.