凤希AI - 场景生成器 - 进Q群与更多的群友学习：775649071

FxAiMultiLineText (凤希AI - 场景生成器 - 进Q群与更多的群友学习：775649071) is the scene-list node of the fxai-toolkit, and the display name is doing double duty: it's a scene generator, and its title literally advertises the author's QQ group. You paste a JSON array of scene descriptions into lines_data, and it turns them into a LIST of prompts plus a computed 循环次数 - a loop count you can wire into a loop node so your graph runs exactly once per scene. It's the "I have a storyboard as text, make it drive my render" node.

The mechanism is simple but the outputs are the point. It parses lines_data (JSON array of strings) into 提示词数据 (LIST - feed this to the grid node or any list consumer). Then it computes the loop count from two optional fields: 开始序号 and 结束序号. Leave both at 0 and 循环次数 equals the number of scenes, starting from index 0. Set 开始序号 to skip the first N scenes; set 结束序号 to bound the run (a value equal to start means "just this one"). The 开始索引 output is the computed starting position (start minus 1), so a loop can pick up exactly where the last run left off. That "loop count + start index + data list" triple is a complete batch contract: the loop reads the count, the list feeds each iteration, and the index keeps resume-correct.

Inputs: lines_data (multiline JSON), 开始序号, 结束序号, and 刷新标记 (a force-input INT that re-triggers evaluation when changed - handy after you edit the list mid-session). Outputs: 提示词数据 (LIST), 循环次数 (INT), 开始索引 (INT).

This node sits in the 凤希AI/图片 (image) category rather than prompts, because in the pack's story workflow a "scene" is the unit of image generation - each scene text gets turned into a grid or single image before the video stage. Pair it with FxAiMultiLineTextGrid (which formats the whole list into one multi-panel "N-grid" prompt) or with a per-index prompt picker inside the loop.

Install: ComfyUI Manager → "fxai-toolkit", or cd ComfyUI/custom_nodes && git clone https://github.com/fxai666/fxai-toolkit, restart. Auto-installs soundfile/psutil; FFmpeg for audio paths. Chinese-first pack from 凤希AI (the title's QQ group is literal - support is in QQ 775649071 and on Bilibili; there's no English Reddit community around it).

Gotchas: a malformed JSON array silently becomes an empty list, and an empty list returns a loop count of 1 (not 0) so the graph still runs once - that's a "never hang forever" design choice that can surprise you as a stray extra pass if your data failed to parse. And the indices are 1-based from the UI's perspective (开始序号 1 = first scene) while the returned 开始索引 is 0-based - keep that straight when wiring into a counter.