HSWQ Save Image

Let's be straight about what this is: a re-skinned version of ComfyUI's built-in Save Image node that adds two widgets the stock one doesn't have. If you're here hoping for something exotic, you can relax - the interesting part of the HSWQ pack is the quantization tooling. This node is the mundane end of the workflow, and that's fine.

What it gives you

• format - PNG (default) or JPG. The stock Save Image is PNG-only; if you want a JPG out the other end, this is the whole difference.
• quality (JPG only) - 1–100, default 95. Exactly what it says: the JPEG quality, used only when format is JPG, ignored for PNG. Since PNG is lossless, a quality value would be meaningless there.
• filename_prefix - same behavior as stock, default ComfyUI. Your files land in the standard output directory with the usual counter suffix.
• images - the IMAGE you're saving.

Under the hood it calls the same folder_paths.get_output_directory() the stock node uses, and for PNG it embeds the workflow prompt and extra_pnginfo into the file's text chunks, which is how ComfyUI workflows travel inside their own output images. Drag the PNG back in and the graph reconstructs - same metadata behavior as stock.

Why it exists

The honest answer: convenience and consistency, not capability. The pack's own workflows end in this node so there's one save path to maintain, and if the author's quantized pipelines ever need a different save behavior, this is the hook. It's an output node, so it has no return socket - it terminates the graph, same as Save Image.

If you want JPG output or per-run quality control without adding a second node pack, this is a perfectly reasonable swap for the stock save node. If you never save JPGs, the stock Save Image does everything you need and there's no compelling reason to switch.

Installing it

It ships with the rest of the HSWQ pack, so install is the pack install:

cd ComfyUI/custom_nodes
git clone https://github.com/ussoewwin/ComfyUI-HSWQ-Loader-and-Tools

Restart ComfyUI. Or use ComfyUI Manager and search the pack title. Worth a heads-up: the pack's requirements.txt is heavy (insightface, onnxruntime, peft, diffusers, transformers - face-detection and quantization deps that exist for the other nodes in the pack). If you're installing the whole pack just for this save node, you're paying for a lot of machinery you won't use. If that's your situation, the stock Save Image is genuinely the better call. If you're running HSWQ workflows anyway, this node is already there - use it.