Meux Multi Save Image

If you've ever run a workflow that produces eight variations and then had to drop eight separate Save Image nodes onto the canvas, you know the exact mess MeuxMultiSaveImage cleans up. It's a terminal output node with sixteen image inputs - plug in a handful of batches, give it one filename prefix, and everything lands in ComfyUI/output as numbered PNGs. It was the pack's original node (v1.0.0 shipped with nothing else), and it's still the one most obviously built to make a working day less tedious.

How it works

Every connected input - all sixteen of images_1 through images_16 - gets flattened into a single list of frames, and each frame is written as its own PNG. Files come out as {prefix}_001_00001.png, {prefix}_002_00001.png, and so on: the first number is the input slot, the second is a counter that bumps when a file already exists, so nothing is ever overwritten. It handles RGBA images (alpha preserved), RGB, and grayscale, and it clamps values before writing.

Two options change the output. resize_to_same plus target_width/target_height force every image to a shared size using lanczos resampling - useful when you're saving a comparison grid later and want uniform dimensions. Filenames get sanitized aggressively (no .., /, : and friends) and the write path is validated to stay inside the output directory, so this node won't write anywhere it shouldn't. That's a small thing until it saves you from a genuinely nasty workflow file.

The one honest caveat

There's a save_individually toggle on the panel, and in the current code (v1.7.1) it doesn't actually change the behavior - the save path already splits every batch frame into its own file regardless. Don't hunt for a "combine into one image" mode here; that's not what this node does. Treat the toggle as future-proofing and ignore it.

Inputs and outputs

• images_1 (required) - plug at least one batch in.
• images_2 … images_16 (optional) - the rest of your outputs.
• filename_prefix - the base name; defaults to ComfyUI.
• save_individually, resize_to_same, target_width, target_height - the options above.

No outputs - it's an output node, which is the point. Its UI result is the list of written files, same as the built-in Save Image.

Installing it

Standard pack install:

cd ComfyUI/custom_nodes
git clone https://github.com/fchangjun/Baidu_Meux_ComfyTools.git
cd Baidu_Meux_ComfyTools
pip install -r requirements.txt

Restart ComfyUI. Or install via ComfyUI Manager by searching "Baidu Meux ComfyTools". The pack's requirements.txt is shared across all its nodes, so you'll pull in transformers, torchvision, realesrgan and scipy even though this node only needs Pillow and torch - the usual custom-node dependency baggage, and the classic source of version conflicts in a shared Python environment (delete and reinstall the venv when things get weird, per the standing advice).

When to bother

Honestly, if you're saving one or two images, the built-in Save Image is fine and this node buys you nothing. Reach for MeuxMultiSaveImage when you're batch-generating - multiple samplers, several LoRA variants, a pass-through of before/after images - and the save node count is spiraling. Sixteen inputs is a lot of headroom; most people never use more than four, and that's okay.