Nodes/Fantastic Loras/Fantastic Plotter Grid Viewer ๐Ÿ”
ComfyUI Node

Fantastic Plotter Grid Viewer ๐Ÿ”

An interactive gallery for your LoRA sweep, without leaving the canvas

By AdudeguymanยทCreated 3 months agoยทUpdated 16 days agoยท 17
Fantastic Plotter Grid Viewer ๐Ÿ”
  • images
    โ—„metadataโ€”โ–บ
    โ—„global_loras_infoโ€”โ–บ
    โ—„fl_archiveโ–บ
    โ—„fl_grid_refโ–บ

    The Image Saver turns a sweep into a flat grid you can keep. The Fantastic Plotter Grid Viewer is the sibling you keep on the canvas: an output node that displays the same comparison interactively - click a cell to zoom it, hide rows and columns you don't care about, tick cells and compare them side by side with captions. It's the terminal display node for the whole Plotter stack, and it's the most comfortable way to actually pick winners from a sweep.

    How to wire it

    You need the individual cells, not the composed grid - you can't pull cells back out of a flattened image. So connect the Image Saver's images (the per-cell list, passed through at full resolution) and metadata outputs, plus global_loras_info if you want the globals strip labeled. The Saver has a ๐Ÿ” Add Grid Viewer (connected) button that wires all of this for you - that's the intended path, and it's what I'd use. The viewer is an output node with no outputs of its own; its required inputs are images (IMAGE) and metadata (STRING), with optional global_loras_info, plus two hidden frontend-managed strings (fl_archive, fl_grid_ref) for archive settings and grid state restoration.

    What you can do with it

    • Zoom - click any cell and it grows into a large centered view with its full metadata (lora, strength, globals); click off it and it shrinks back.
    • Hide rows and columns - every header has a โœ•, so you can strip out the losers and focus on a subset without re-running the graph. Hidden ones come back as chips or via Reset filters.
    • Compare - tick two or more cells, hit Compare (N), and they line up side by side, each captioned with exactly what the Plotter used. From there you can export the side-by-side as one PNG, or ๐Ÿ’พ Save comparison to bank that set of cells under a name and reopen it later.

    The grid persists with your workflow - reopen the workflow or switch tabs and the last run's grid comes back (only lightweight image references are stored, not pixels). A saved comparison is bound to its run's grid; running the Plotter again is a clean slate, which is by design since the old grid no longer exists.

    Grids on disk

    By default grid images live in ComfyUI's temp folder, which clears on restart - so the in-workflow grid restores across tab switches but not a full restart. For permanent copies, save grids to disk: ๐Ÿ’พ Save Grid pins a grid under output/fantastic-loras-grids/<run_id>/ with a manifest.json (kept until you delete it), or turn on auto-save every run in Archive Settings. Auto-saved runs are subject to cleanup - delete runs older than N days (default 14) and/or keep only the last N runs - while pinned grids are exempt. Your Archive Settings are remembered as global defaults in user/fantastic-loras/archive_defaults.json, so new viewers start from your preferences. One caveat: saved grids reference files by name, so moving the workflow to another machine won't bring the grids with it.

    Install

    cd ComfyUI/custom_nodes
    git clone https://github.com/Adudeguyman/comfyui_fantastic-loras
    

    or ComfyUI Manager โ†’ search Fantastic Loras, restart, hard-refresh. No dependencies.

    Common issues

    The gotchas are mostly about expectations. If a saved grid won't reload, it's the file-reference caveat above - the output files moved or were cleaned. If the grid seems to "forget" a run, that's either the temp-folder clearing on restart (save or pin it) or the clean-slate rule (a new run legitimately replaces the old grid). And the usual pack fix for anything rendering wrong after an update is a hard browser refresh first, then a ComfyUI restart. It's a display node - it can't break your generation, only your view of it - which makes it a low-risk add to any Plotter workflow.

    Categoryloaders

    Inputs (5)

    NameTypeDefaultDescription
    imagesIMAGEโ€”
    metadataSTRINGโ€”
    global_loras_infooptSTRINGโ€”
    fl_archiveoptSTRINGโ€”
    fl_grid_refoptSTRINGโ€”

    Outputs (0)

    No outputs