Nodes/comfyui-mini-nodes/mini_image_save
ComfyUI Node

mini_image_save

Exporting a PNG that does NOT carry your workflow

By catmaxzj·Created 4 months ago·Updated 2 days ago· 28
mini_image_save
  • 图像
    文件名头%y%m%d_%H%M
    保存工作流true

    In ComfyUI, every PNG that leaves the stock SaveImage node carries the entire workflow that produced it - the node graph, the seeds, every widget value - embedded as text chunks inside the file. That's the culture: drag the image back onto the canvas and the whole pipeline reconstructs. It's also a problem when you don't want that. Sharing a WIP to a client, posting a comparison where the settings are private, or just shipping a file where the embedded graph is noise - the stock node has no switch for "write the pixels, skip the recipe."

    mini_image_save is that switch. It's a drop-in replacement for SaveImage with one extra boolean: 保存工作流 (save workflow), on by default, and when you turn it off the PNG goes out with no embedded graph at all.

    How it works

    It's a thin output node - same pattern as the stock save, no model, no dependencies beyond what ComfyUI bundles. It writes a PNG into your output folder via ComfyUI's own save-path machinery. With 保存工作流 on, it embeds both the prompt and extra_pnginfo chunks, exactly like the stock node; with it off, those chunks never get written. You get a clean image file, and the pixels are identical either way.

    There's a second small win hiding in the filename handling. The 文件名头 (filename prefix) field accepts strftime-style date/time tokens:

    • %y %m %d %H %M %S - two-digit year, month, day, hour, minute, second
    • %date%2026-08-25, %time%14-30-22

    So the default %y%m%d_%H%M produces names like 260825_1430_01.png, and batches get a two-digit counter (01, 02, …) so a 12-image batch sorts correctly instead of going 1, 10, 11, 2. It also sanitizes the prefix through os.path.basename, which quietly kills any path-injection attempt - a nicety the stock node doesn't bother with.

    The inputs that matter

    There are only three, and the node has no outputs (it's an output node - results show up in the UI preview and the output directory):

    • 图像 (image) - the IMAGE tensor to save.
    • 文件名头 (filename prefix) - the string, with the tokens above. Default %y%m%d_%H%M.
    • 保存工作流 (save workflow) - boolean, default true. This is the whole reason the node exists.

    One honest warning for English speakers: the input names are in Chinese. 图像 = image, 文件名头 = filename prefix, 保存工作流 = save workflow. You'll get used to it in about two runs, but it trips people the first time.

    Install

    ComfyUI Manager → search "comfyui-mini-nodes" → install, or:

    cd ComfyUI/custom_nodes
    git clone https://github.com/catmaxzj/comfyui-mini-nodes
    

    Then restart. No requirements.txt, no model files, nothing heavy.

    When to actually use it

    Honestly, most days you keep 保存工作流 on - the embedded workflow is how you revisit what you made and how you share it properly. Turn it off when you're handing a file to someone who doesn't need your graph (or who you don't want peeking at it), or when you're uploading somewhere that re-encodes anyway - Reddit and most CDN thumbnails strip the metadata the moment they touch it, so carrying it through those hops is wasted effort. One caveat: this node writes PNG only, same as the stock save. No JPEG, no WebP - if you need those, a format-conversion node is the next hop in the graph.

    The "no workflow" export is the feature. Everything else is just a competent save node that happens to come in a four-node pack.

    Categorymini_nodes

    Inputs (3)

    NameTypeDefaultDescription
    图像IMAGE
    文件名头STRING%y%m%d_%H%M
    保存工作流BOOLEANtrue

    Outputs (0)

    No outputs