凤希AI - 图像管理器 - 群友扫地僧友情参与开发
Upload, browse, reorder, and auto-number your assets
- 图片
- 文件列表
- 文件夹路径
- 图片总数
- 图片
- 刷新标记
ComfyUI's default way of getting images in - drag a file into Load Image - stops scaling the moment you're managing dozens of assets for a segmented video shoot. You want a library: a folder you can drop images into, see them at a glance, reorder them, and have them come back in a predictable order. That's what FxAiImageManager is: a folder-backed image library that lives inside ComfyUI, with a browser panel, drag-reordering, and automatic sequential filenames so "frame order" is never ambiguous.
Its display name has a fun detail worth knowing: "群友扫地僧友情参与开发" - a member of the author's community, going by the nickname "Sweeping Monk," helped build it. That tells you a lot about where this pack comes from: it's a Chinese-community project (凤希AI, fxai.site), iterated on by its own users in a QQ group, and this node is a literal group effort.
How it works
Images live on disk under ComfyUI/fxai/image/<目录> - the 目录 input (default sucai, which is "素材", meaning "assets") selects which subfolder you're managing. The node registers HTTP routes on ComfyUI's own server (/fxai/image/preview, /list, /upload, /apply, ...) that power a JS frontend panel: upload images through the browser, preview them, drag to reorder, and the server renames files to 000.png, 001.png, 002.png... in the new order. Auto-numbering means the loaders downstream can count on lexical order.
The node itself does two things in the graph. If you feed an 图片 tensor into it, it saves those frames into the folder as numbered PNGs (using the next free number) - so you can route generated frames straight into your asset library. And it reports the library state: 文件列表 (newline-joined filenames), 文件夹路径, 图片总数, plus a passthrough 图片 and a 刷新标记 that you can bump to force the panel to reload.
Inputs and outputs
- 目录 - which library subfolder (default
sucai). - 图片 (optional) - feed frames in to save them to disk.
- 刷新标记 (optional) - a number to trigger a refresh.
Outputs: 文件列表 (STRING), 文件夹路径 (STRING), 图片总数 (INT), 图片 (IMAGE passthrough), 刷新标记 (INT).
Installing it
Part of fxai-toolkit (凤希AI, MIT). Install via ComfyUI Manager (search "fxai-toolkit") or:
cd ComfyUI/custom_nodes
git clone https://github.com/fxai666/fxai-toolkit
# restart ComfyUI
No models. It needs ComfyUI's web server (it's all HTTP routes), so it won't do anything useful in a headless API-only setup. The pack auto-installs soundfile and psutil at first load. Chinese labels; support via QQ group 775649071 and Bilibili. There's essentially no English-language community documentation for this pack, so treat the author's own channel as the source of truth.
Where people get burned
The path is the gotcha: files live under your ComfyUI install root (in fxai/image/), not in a folder of your choosing - the 目录 input only picks a subfolder within that library. If you were expecting it to read an arbitrary folder you already have, it won't; use the pack's folder-path loaders for that. Also, reordering is a server-side rename operation, so if the ComfyUI server is running, files are genuinely renamed - don't do the same reorder from two browser tabs. And the preview/list routes have no auth by default, which is normal for ComfyUI but worth remembering if you expose your instance to a network (see the ComfyUI_LLMVISION incident for why that's a real concern).
Inputs (3)
| Name | Type | Default | Description |
|---|---|---|---|
| 目录 | STRING | sucai | — |
| 图片opt | IMAGE | — | |
| 刷新标记opt | INT | — |
Outputs (5)
| Name | Type | Description |
|---|---|---|
| 文件列表 | STRING | — |
| 文件夹路径 | STRING | — |
| 图片总数 | INT | — |
| 图片 | IMAGE | — |
| 刷新标记 | INT | — |