Extract Lora Name
The tiny node WextraUI already retired (use this instead)
- STRING
First, the thing nobody tells you on this page: "Extract Lora Name" (LoraName) is retired. The WextraUI changelog moved it to src/legacy/ on 4 September 2026, when the pack renamed everything to the W prefix. Current installs no longer register the class, so if you install WextraUI today and search for it, it simply won't be there. You've landed here from an old workflow or an old blog post. Good news: what it did is still easy to get, just in a different place.
What it actually did
It's a one-trick string node. Feed it a LoRA path and it hands back a clean name: it splits the input on split_char, keeps the last segment, then cuts at the first . to drop the extension.
loraName(STRING) - the path, e.g.models/loras/my_character_v2.safetensorssplit_char(STRING, default\) - what separates folders in your path- Output: one STRING with the clean name, e.g.
my_character_v2
Its natural home was the filename game: you want a saved image or a run tag to say which LoRA was active, and LoRA filenames are usually the most memorable label you have (LoRAs carry their trigger words and identity in the file name far more reliably than anywhere else). Wire the output into WSave Image's filename parts and your output folder becomes self-documenting.
The gotchas that will bite you
The default split character is a backslash. That's Windows muscle memory, and it's the trap. On Linux or macOS, or anywhere a path uses forward slashes, models/loras/foo.safetensors with the default \ doesn't split at all - you get back models/loras/foo, folder included. Set split_char to / and it's clean. If your filename has no folder at all, the default happens to be harmless.
It cuts at the first dot, not the last. add_detail.1.0.safetensors becomes add_detail - it eats version suffixes like v2 when the dot lands mid-name, and that's often exactly the part you wanted to keep for a filename.
I ran the shipped code to confirm both, and that's the whole mechanism. No API, no model files, nothing clever.
What to use instead
The pack's successor node, WLoad Lora & Trigger (wxLoraLoaderTrigger), does this properly as one of its outputs (name). It normalizes both slash styles and splits on the last dot, so add_detail.1.0.safetensors survives intact. And since it loads the LoRA and collects its trigger words in the same node, you get the name for free while doing something useful. If you only need a label for a caption or a note, honestly typing it, or grabbing the base filename from any standard LoraLoader widget, beats resurrecting a retired node.
Install
If you do want the pack: ComfyUI Manager → search WextraUI, or:
cd ComfyUI/custom_nodes
git clone https://github.com/sickbraintwo/WextraUI
No extra dependencies - it's a plain Python pack. Just note again: the current release won't give you this specific node. To run LoraName you'd have to pin a commit from before 4 September 2026 or manually register the copy in src/legacy/, which I'd only bother with if you're restoring an old workflow exactly. Everyone else: point at WLoad Lora & Trigger and move on.
Inputs (2)
| Name | Type | Default | Description |
|---|---|---|---|
| loraName | STRING | — | |
| split_char | STRING | \ | — |
Outputs (1)
| Name | Type | Description |
|---|---|---|
| STRING | STRING | — |