PD_rename_image

Downloaded 500 reference images that need to become img_0001.jpg … img_0500.jpg, capped at 1024px on the long edge, all in JPG. That's the job this node was built for. It's a one-folder batch tool: rename to a numbered scheme, resize keeping aspect ratio, and convert between JPG and PNG, with a single report string coming back out. If you've ever scripted this with PIL by hand, this is that script, wrapped in a ComfyUI node.

It overlaps with PD_number_star (also in this pack), but the focus here is different: that one is a general renamer with format conversion bolted on; this one is specifically "images → numbered, resized, converted." Pick by which defaults match your task.

How it works

It lists the image files in the folder (jpg/jpeg/png/bmp/tiff/webp), sorts them by name so the numbering order is deterministic, then for each one:

• Builds the new name from rename_pattern. The default img_{index:04d} gives img_0001, img_0002, …; {index} is the 1-based position, and {original} is the original stem if you want it preserved.
• Converts to output_format - JPG flattens alpha onto a white background, PNG converts palette/LA modes to RGBA.
• Resizes the long edge down to max_size (LANCZOS), only if the image is actually larger.
• Saves the result and deletes the original file if the name changed.

So it's not just a rename - it's a re-encode-and-replace. Alpha loss in JPG is the main data casualty to be aware of.

The inputs that matter

• input_path - the folder. Must exist and be a directory, or you get an error string back.
• rename_pattern - img_{index:04d} style; supports {index} with zero-padding via format spec and {original}.
• max_size (128–4096, default 1024) - longest-edge cap.
• output_format - JPG or PNG.

Output: result - a text report with path, size cap, format, and per-file success/failure counts.

Installing it

Part of Comfyui_PDuse by 7BEII. ComfyUI Manager (search "PDuse") or:

cd ComfyUI/custom_nodes
git clone https://github.com/7BEII/Comfyui_PDuse.git
cd Comfyui_PDuse
pip install -r requirements.txt

Restart ComfyUI. No models; pack deps are numpy, Pillow, scipy, opencv-python, torchvision. Docs are Chinese-first.

Where people get burned

Originals are deleted. When the new filename differs from the old one, the source file is removed - there's no keep-both option, and unlike PD_RenameV2 (same pack), there's no copy-to-a-new-folder escape hatch here. Work on a copy. Second, the numbering order comes from os.listdir + sort, which is fine for img_0001.jpg style names but will renumber everything if you rerun it on a folder that's already been renamed - rerunning is not idempotent. Third, JPG conversion destroys transparency; if your set has PNGs with alpha, don't pick JPG unless you're okay with white backgrounds baked in.