ComfyUI-lora-FindingLora Delivers Quick LoRA Finding And Stacking
The ComfyUI-lora-FindingLora custom node replaces ComfyUI’s stock LoRA loader with fuzzy search, bookmarking, trigger word storage, and one-click LoRA stacking. It allows you to find any LoRA from a large collection by typing a few characters, and it remembers your favorites for quick access. A built-in chaining button automatically creates and wires a second loader into the model chain, eliminating manual rewiring when you want to apply multiple LoRAs.
The node was created by developer Shootthesound (who also developed ComfyUI-Wan-I2v-Control and ComfyUI-ReferenceLatentPlus) to address frustrations he experienced with the default loader. He got tired of scrolling endless dropdowns, manually rewiring nodes to stack LoRAs, and trying to recall trigger words. So he combined everything he wished the stock loader offered—bookmarks, trigger storage, fuzzy search, and auto-chaining—into a single, efficient interface.
Bookmarks, trigger words, and auto-chaining
Key Features
- Fuzzy search ranks substring matches higher.
- One-click bookmark saves any LoRA.
- Store trigger words and copy with one click.
- Button spawns and auto-wires another loader.
- Modal picker replaces clumsy dropdowns.
- Bookmarks sync live across every node.
- Click trigger row to copy phrase instantly.
- Model-only loader for Flux, Klein, and Wan.
ComfyUI users who juggle hundreds or thousands of LoRAs will find the bookmark and search system drastically cuts down time spent hunting for the right file. Small creative studios and privacy-focused professionals can work entirely locally, with trigger words stored and pasted directly into prompts without external tools. The one-click chaining feature makes experimenting with LoRA combinations faster and less error-prone, especially during iterative design work.
Developer notes and future plans
Bookmarks are saved as a plain JSON file that you can open and edit for bulk imports, making backup or sharing easy. The built-in search uses two scoring tiers—contiguous substring matches always appear before scattered character matches, so results stay predictable. Because the node is a model-only loader, it works without changes on modern pipelines like Flux, Klein 9B, and Wan, where CLIP isn’t part of the LoRA chain.
"I haven't seen another LoRA-loader pack do this combination — bookmarks + trigger storage + click-to-copy + fuzzy search + auto-chaining — so here it is." — Source: GitHub
Project Links
