👑 FRED Extract Lora parameter

Your workflow applied a LoRA, but by the time you're looking at the output you have no idea which one or at what weight. FRED_ExtractLora_parameter fixes that by reading the LoRA configuration out of whatever loader or stacker node your model wire passes through - up to ten LoRAs, names cleaned of paths and extensions, weights included, all as plain outputs.

What it does

Single input, lora_stack (type *). Like the KSampler extractor in this pack, the node doesn't consume the data - it inspects the node the wire comes from, finds its LoRA settings, and emits:

• lora_name_1 through lora_name_10 (STRING) - bare filenames, e.g. my_char_lora rather than loras/my_char_lora.safetensors.
• lora_weight_1 through lora_weight_10 (FLOAT) - the strength each was applied at.
• help, the standard card.

Unused slots come back as empty name / zero weight, so the output shape is stable even if you only stack three.

The IS_CHANGED behavior is worth a mention: the node is marked to always re-run on each execution, which is deliberate - LoRA configs change constantly and a cached stale read would defeat the entire purpose.

Where it shines

The headline use is filename honesty. The FRED Image Saver's token system has %lora_name_1 / %lora_weight_1 and friends, but only if those values arrive on the saver's inputs. Wire these extracted names and weights into the saver and your filenames reflect the real stack - my_char_0.8_20260816.png instead of a generic counter. When you're comparing ten variants of a character at different weights, that naming is the difference between a usable archive and a mystery pile.

It's also the bridge for anything that wants LoRA info as plain strings: log files (via FRED_Save_Text_File), your own text display nodes, or a workflow that routes differently based on which LoRA ran.

What it reads

It understands the standard LoRA loader/stacker convention - nodes that expose a lora_name_N + lora_weight_N (or strength) pattern, including the FRED Auto LoRA loaders and common third-party stackers. Connect it to a LoraLoaderModelOnly or a custom node with unusual field names and you may get fewer slots filled; the extraction is pragmatic, matching the conventional names.

Install

cd ComfyUI/custom_nodes
git clone https://github.com/Poukpalaova/ComfyUI-FRED-Nodes_v2.git

Restart, or grab "ComfyUI FRED Nodes v2" from ComfyUI Manager. Base pack dependencies only.

Gotchas

Wire it after the last LoRA node in your chain if you want the final composed state - if you attach it before the stacker, it only sees the LoRAs upstream of the tap point. And remember the extracted values reflect what's configured, which is exactly what you want for metadata; it doesn't magically know about a LoRA that was bypassed, because a bypassed slot reads as name-empty anyway.