🗃️ Batch Image Saver

Most save nodes make you pick a filename prefix and then every image becomes my_workflow_00001.png, _00002, and so on. Fine if you're generating a stack of fresh images. Useless if you're processing a folder of existing files and you want the result to keep the original name so you can match them back up later.

That's the niche the 🗃️ Batch Image Saver lives in. It's the natural partner to this pack's Folder Image Loader: you point the loader at a directory, run your pipeline, and feed the loader's filename output into this node's 原文件名 input. Every output image is then saved under its source name - just with the extension swapped. You get dragon_pose_a.png → dragon_pose_a.jpg instead of img_00042.jpg. That one wire turns a batch-processing workflow from a sorting nightmare into something you can actually manage.

What it saves (and what it saves alongside)

Three optional things, independently:

• Images - PNG, JPG, or WebP via the 文件格式 dropdown. The 质量 slider (1–100) only matters for JPG/WebP; PNG ignores it.
• Text - wire the text input and it writes a same-named .txt next to each image. Wire json and you get a .json sidecar. The tooltips hint where this came from: the json input is meant to be fed from Ideogram4 Text Encode's json_text output, so every image ships with its structured prompt next to it.
• Workflow metadata - the 嵌入工作流 toggle is on by default, which means your prompt and graph are baked into PNG text chunks just like core Save Image does. Flip it off for client-facing delivery, or when you don't want your workflow leaking out. Note it's PNG-only: JPG and WebP never carry it, so if you save JPG you lose the drag-back-into-ComfyUI property entirely.

The 覆盖已有文件 toggle decides collisions. On (default) means a same-named file just gets overwritten - which is what you want when you're re-running a batch with the original names connected. Off means it auto-appends _1, _2, … so nothing is ever clobbered.

The 输出文件夹 accepts an absolute path, or a relative one that resolves against ComfyUI's output directory. Leave images unwired and it'll happily save only the text sidecars - handy when you're turning a folder of prompts into files.

Installing it

It's part of the Louis Use pack, so you get all the other nodes here at once:

cd ComfyUI/custom_nodes
git clone https://github.com/LouisLU1997/ComfyUI-louis-use Louis_use

Or search "Louis Use" in ComfyUI Manager. Restart ComfyUI after either. No extra Python packages needed - this node runs on the torch/Pillow/numpy that ship with ComfyUI.

Where people get burned

• Empty folder path throws. The node raises if 输出文件夹 is blank, and relative paths resolve against the output dir, so batch_out isn't a folder in your home directory.
• JPEG drops the workflow. If you switch to JPG for size and then wonder why the file won't drag back into ComfyUI, that's the 嵌入工作流 note biting you - it's not a bug, it's the format.
• Filename sync needs the loader's filename output, not the folder path. The input is a list of strings, one per line. If you're getting img_00001 style names, that wire isn't connected.

It's a simple node - a folder, a format, an overwrite toggle - but the original-filename sync is the feature that makes it worth hunting down, and nothing else in core does it.