凤希AI - 场景多宫格加载器

FxAiMultiLineTextGrid (凤希AI - 场景多宫格加载器) is the formatting node that turns a list of scene descriptions into a single prompt for a multi-panel image. "多宫格" means multi-grid - the contact-sheet / storyboard layout where one image holds several panels, each panel a different scene. You hand it the 提示词数据 LIST (the same LIST that FxAiMultiLineText outputs), and it builds one prompt that describes an N-panel grid with every scene in order. The model gets: "N-grid image, multi-panel storyboard layout," then scene 1, scene 2, and so on, each labeled 场景N:.

This is one of those nodes whose whole value is in the exact wording it generates, and the wording is sensible: {N}宫格画面，多格分镜布局。 - "N-grid picture, multi-panel storyboard layout" - followed by the optional 通用提示词 prefix and 尾部通用提示词 suffix, then each scene on its own line. The prefix/suffix are the same convention as FxAiMultiLinePrompt: put your global style and character anchors once in 通用提示词, per-scene variety in the list, shared closing instructions in 尾部通用提示词. If the list is empty it degrades to just prefix+suffix rather than erroring.

Inputs: 提示词数据 (LIST, force-input), 通用提示词, 尾部通用提示词 (optional strings). Output: a single 提示词 STRING - one prompt, ready for a text encoder, aimed at a 2×2 or 3×3 grid generation. It does not touch images itself; it only writes the prompt that describes the grid.

Where it fits in the pack: it's the bridge between the storyboard layer and image generation. FxAiMultiLineText manages the scene list and loop bookkeeping; this node formats a whole list into a contact-sheet prompt (good for previewing a storyboard as one image, or generating a character sheet), while per-scene generation would walk the list one index at a time instead. If you're generating one image per scene, you don't want this node - you want the per-index path. If you want to see all scenes at once as panels, this is exactly it.

Install: ComfyUI Manager → search "fxai-toolkit", or cd ComfyUI/custom_nodes && git clone https://github.com/fxai666/fxai-toolkit, restart. Pack auto-installs soundfile/psutil; FFmpeg for audio paths. Chinese labels throughout (凤希AI builds for a Chinese-speaking community - QQ 775649071, Bilibili; no English Reddit presence for the pack).

The practical note: grid prompts are one of the least reliable generations in diffusion - models regularly draw 3 panels when you ask for 4, or fill two panels with the same scene. That's a model limitation, not this node's bug; keep the scene count small (4–6) and make each scene line distinct enough that the model has something to hang a panel on. If your grid comes back scrambled, that's the model, and you'll be happier generating scenes one at a time anyway.