凤希AI - 视频场景管理V2 - 进Q群与更多的群友学习：775649071

FxAiSceneManagerV2 is the upgraded scene database for the pack's long-video flows, and the upgrade is all about audio. V1 tracked a scene's duration, prompt, and audio index. V2 adds 音频开始 (audio start time) as a first-class field - because in real film work, a scene's audio doesn't always start at second zero. It also adds a third output, 分段时长 (segment durations as a LIST), which makes it trivial to drive frame calculations and loop control.

The mechanism is the same tolerant normalization as V1, with richer fields. You paste JSON into the lines_data multiline box; a list of lists maps positionally as [音频时长, 提示词文本, 音频索引, 音频开始, 图片索引, 尾帧位置, 转场]. Note the field order changed from V1 - field 0 is now 音频时长 (audio duration, default 15.0) and field 3 is 音频开始 (audio start in seconds, default 0.0). Each row becomes a dict with 序号, 音频时长, 提示词文本, 音频索引 (default 0), 音频开始, 图片索引 (default -1), 尾帧位置 (default -1), and 转场 (default 1). If you paste an empty list, it doesn't error - it defaults to a single 15-second empty scene, which is a nice "no crash, you'll see" behavior. Plain strings still work as prompt-only scenes.

Outputs: 总行数 (INT), 场景数据 (LIST - feed this to FxAiSceneLoadV2), and 分段时长 (LIST of floats, one per scene's audio duration). That last one is the killer feature: you can feed it straight into the pack's frame-calculator nodes so total frames stay in sync with the audio timeline, instead of recomputing durations downstream.

Two things worth flagging. First, the field-order difference from V1 will bite you if you copy a V1 JSON: a V1 row like [5.0, "prompt", 0, -1, -1, 1] maps to different fields in V2 because of the inserted 音频开始. If a scene suddenly has a 0.0-second audio start or a shifted prompt, that's why - regenerate the JSON for V2. Second, like V1 it's a tolerant parser, so malformed JSON becomes the default single scene rather than an error; check 总行数 before you trust it. It has its own front-end editor in the 凤希AI menu (the display name points users at the author's QQ group for support).

It's in 凤希AI/场景管理, part of fxai-toolkit (凤希AI / fxai.site):

cd ComfyUI/custom_nodes
git clone https://github.com/fxai666/fxai-toolkit

or ComfyUI Manager → search "fxai-toolkit", then restart. If your scenes sync to a soundtrack, V2 is the one you want - V1 can't express "this scene starts 3 seconds into the audio."