Faishme Load Image from Glob
Load a Whole Folder of Images at Once — and Keep the File Paths
- IMAGE
- FILE PATH
- width
- height
Faishme Load Image from Glob is the pack's bulk image loader, and it does two things the built-in Load Image won't: it pulls in an entire directory (or glob match) of images in one go, and it hands back the file path for every image it loaded. That second bit is the feature - it means you can later save generated output back to the same folders with Faishme Save Image, which is exactly how this pack's fashion pipeline is meant to be wired.
How it works
Give it a pattern like C:/shots/*.png and it runs a glob, filters to jpg, jpeg, png, and webp, sorts them, and loads every match in parallel via a thread pool. Each image is resized to your width × height unless you tick default_size, which keeps each file's original dimensions instead. It then outputs the loaded IMAGE batch, a parallel list of FILE PATH strings, and the width/height it used.
There's a second input path that's easy to miss: file_path. The node is list-aware, so if you wire in a list of explicit paths (from another node, say), it loads those instead of globbing. The glob is the fallback when file_path is empty. Either way, a pattern with no matches raises a FileNotFoundError rather than silently doing nothing - so if your batch comes back empty, check the pattern first.
Inputs that matter
- pattern - the glob string, e.g.
C:/ComfyUI/output/input/*.jpg. This is the one you'll actually set. - width / height - defaults are 1024 × 1536, i.e. portrait. That's a giveaway about the fashion use case, and a trap if you're loading landscape images you wanted at original size - set default_size to true and the resize is skipped entirely.
- default_size - when true, ignores width/height and keeps each image's native size.
Outputs: IMAGE (a list, one tensor per file), FILE PATH (the same length, matching order), plus width and height as plain ints.
Wiring it with the rest of the pack
The natural pairing: load a folder with this node, run your generation, then send the new images plus the original FILE PATH list into Faishme Save Image, which writes each output back next to its source. That round-trip (read folder → process → write beside source) is the whole point of the design, and it's why FILE PATH is an output at all.
Installing it
Same as every node in this pack: search "ComfyUI_faishme" in ComfyUI Manager, or:
cd ComfyUI/custom_nodes
git clone https://github.com/AkashKarnatak/ComfyUI_faishme
Restart ComfyUI and it appears under FaishmeNodes.
Where people get burned
Three things, in order of likelihood: a pattern with no matches (you get the FileNotFoundError above); forgetting that width/height resize applies by default, silently re-fitting every image to 1024×1536; and assuming the IMAGE output is a single batched tensor when it's actually a list - which matters if you feed it into a node expecting one tensor. Otherwise it's one of the more genuinely useful nodes in the pack, and the parallel loading keeps big folders from feeling slow.
Inputs (5)
| Name | Type | Default | Description |
|---|---|---|---|
| pattern | STRING | — | |
| file_path | STRING | — | |
| width | INT | 10240–8192 | — |
| height | INT | 15360–8192 | — |
| default_size | BOOLEAN | false | — |
Outputs (4)
| Name | Type | Description |
|---|---|---|
| IMAGE | IMAGE | — |
| FILE PATH | STRING | — |
| width | INT | — |
| height | INT | — |