Load Image From Output Subdirectory With Workflow Ovum

Same job as LoadImageFromOutputWithWorkflowOvum, one extra knob: this node loads an image from a subdirectory under your output folder and pulls out its embedded prompt/workflow metadata. If your saver is set up to write into organized subfolders - output/faces/, output/batch_2026_08/ - this is the loader that matches your file layout instead of forcing everything into one flat list.

It's the round-tripping sibling: browse the output folder, jump into a subdirectory, pick a file, and get back the image, its mask, its filepath, the embedded PROMPT&WORKFLOW dict, and the pack's IMAGE_EX bundle. Both nodes carry the EXPERIMENTAL = True flag in the source, so expect occasional rough edges, not a full rewrite of how ComfyUI loads images.

Inputs and outputs

• image - the file picker COMBO, pointing at the output folder (image_folder: "output"), with upload support, a refresh button, and control_after_refresh so new files appear.
• output_subdir (optional STRING) - the subdirectory to look in, e.g. my_subdir/nested. Leave it blank and it behaves like the flat version, reading from the top of the output folder. The value gets normalized (leading/trailing slashes stripped) and joined onto the output directory.

Five outputs, identical to the flat variant:

• IMAGE - the loaded image.
• MASK - the image's alpha channel as a mask.
• FILEPATH (STRING) - the resolved full path.
• PROMPT&WORKFLOW (DICT) - the embedded prompt/workflow metadata.
• IMAGE_EX (DICT) - the pack's image-plus-metadata bundle for IMAGE_EX-aware nodes downstream.

Why you'd prefer this over the flat loader

If you've got more than a handful of saved outputs, a flat dropdown becomes a wall of near-identical filenames. Subdirectories give you a way to organize by project or batch, and this node lets you drill into them. The output_subdir field is a plain string - you can even build it dynamically with a format node ("batch_{run_id}") if you're generating per-run folders, which makes this a natural fit for automated pipelines that save into dated or numbered subfolders.

Install and gotchas

Ships in comfy-ovum:

cd ComfyUI/custom_nodes
git clone https://github.com/sfinktah/comfy-ovum

or ComfyUI Manager → comfy-ovum, restart. No models, no keys.

Two things to remember: images saved by tools that strip metadata will come back with an empty PROMPT&WORKFLOW rather than an error (no recipe, but the pixels are fine), and a typo in output_subdir just means the file isn't found - the node resolves the path under the output directory, so if you can't see your files, check the subdir string for stray spaces or wrong casing. It's marked experimental, so keep the plain core Load Image handy as a fallback.