Nodes/LLS-node/LLS Save Image
ComfyUI Node

LLS Save Image

Save image, save mask, or both — with your whole generation pipeline in the PNG metadata

By Gin3601·Created 3 months ago·Updated 3 months ago· 0
LLS Save Image
  • image
  • mask
    filename_prefixLLS
    output_modesave
    save_metadatatrue
    prompt_info
    latent_info
    sample_info
    decode_info
    upscale_info

    Save Image is a node everyone takes for granted until they need to save a mask, or they dig up a week-old PNG and have no idea what produced it. LLS Save Image does both halves: it can save an image, a mask, or both at once, and it stitches the whole LLS pipeline's metadata into the PNG - family, checkpoint name, VAE, seed, prompts - so the file is self-documenting.

    The masking half is the headline feature, because ComfyUI's built-in Save Image doesn't save masks at all. In the pack's repair workflow you're always juggling an original image and the mask that drives the edit, and being able to save them together means your inpainting project is reproducible weeks later. Wire up image and mask, and in save mode the mask lands in the same folder under <filename_prefix>_mask. Neither port is required - connect whichever you have - but if you connect both, both get saved.

    The inputs

    • filename_prefix - the filename stem (default LLS).
    • output_mode - save or preview_only. The latter is a nice touch: preview without committing a file.
    • save_metadata - toggle the PNG metadata writing.
    • Optional prompt_info, latent_info, sample_info, decode_info, upscale_info - the JSON strings emitted by the LLS pipeline nodes. Wire the ones you have: LLS Simple Prompt Encode's prompt info, LLS Simple Empty Latent's latent info, LLS Simple KSampler's sample info, and so on. The node merges them into one coherent metadata block.

    Where it fits

    Every LLS workflow ends at this node, and the value compounds as you add more of the LLS chain: each node you wire in contributes another field of metadata. For beginners the payoff is "drag the PNG back into ComfyUI and the whole workflow reconstructs" - the standard ComfyUI sharing superpower, with your seed and settings preserved. The KB's image-io-metadata essay is the reference on how this property makes workflows self-healing.

    Installing it

    cd ComfyUI/custom_nodes
    git clone https://github.com/Gin3601/LLS-node
    # restart ComfyUI
    

    Or ComfyUI Manager → search "LLS-node". It reuses ComfyUI's native SaveImage under the hood, so no extra dependencies.

    Gotchas

    The preview_only mode is a trap if you forget it's on and wonder where your files went - check output_mode before assuming the save failed. And the info inputs are plain JSON strings with forceInput sockets, so they're meant to be wired from the LLS pipeline nodes rather than typed by hand. If you don't connect them, save_metadata still works - it just records less. This is the node to reach for when you're done iterating and want the receipts.

    CategoryLLS/Image

    Inputs (10)

    NameTypeDefaultDescription
    filename_prefixSTRINGLLS
    output_modeCOMBOsave2 options: save, preview_only
    save_metadataBOOLEANtrue
    imageoptIMAGE
    maskoptMASK
    prompt_infooptSTRING
    latent_infooptSTRING
    sample_infooptSTRING
    decode_infooptSTRING
    upscale_infooptSTRING

    Outputs (0)

    No outputs