Lora Pool (LoraManager)

Lora Pool is the first piece of LoRA Manager's randomizer subsystem - the set of nodes for exploring LoRA combinations instead of pinning down one. On its own it doesn't randomize anything; it defines the bag you draw from. You configure a collection of LoRAs here, and the Randomizer or Cycler nodes pull from that pool according to their own rules.

Why split it out? Because a pool is reusable. Define your "anime styles" pool once, and feed it into a Randomizer for one experiment and a Cycler for another. It keeps the "which LoRAs are candidates" decision separate from the "how do I pick among them" decision, which is exactly the kind of clean separation that makes the randomizer worth using for style exploration and batch variety.

There's a practical reason this matters for LoRAs specifically. Interactions between stacked LoRAs are genuinely unpredictable - the standard advice is to add them one at a time with a fixed seed because you can't guess how two will blend. A pool turns that manual poking into something you can automate: assemble a set of plausible candidates once, then let the Randomizer or Cycler explore the combinations for you instead of rewiring the graph every time.

How it works

The node holds a pool_config - a widget-driven configuration where you assemble the LoRAs (and their weight settings) that make up the pool. It emits that as a POOL_CONFIG output, which is the standard connector the other randomizer nodes accept on their optional pool_config input. Nothing gets applied to a model here; the pool is a definition that flows downstream.

Inputs and outputs that matter

• pool_config (LORA_POOL_CONFIG, required) - the node's own configuration widget where you build the pool of candidate LoRAs.

One output: POOL_CONFIG - wire it into a Lora Randomizer or Lora Cycler's pool_config input.

How to install it

Via ComfyUI Manager: search lora-manager, Install, restart. Or manually:

cd ComfyUI/custom_nodes
git clone https://github.com/willmiao/ComfyUI-Lora-Manager.git
cd ComfyUI-Lora-Manager
pip install -r requirements.txt

then restart. Light Python install.

Common issues & troubleshooting

Nothing happens when I run just this. Correct - a pool is a definition, not an action. You need a Randomizer or Cycler downstream to actually pick from it and produce a LORA_STACK a loader can apply.

Only some pool LoRAs ever show up. That's the randomizer or cycler doing its job - it selects a subset per run. If you want everything at once, you don't want a pool at all; use a plain Lora Stacker.

Base-model mismatch. A pool can hold LoRAs from different architectures, but a draw that lands on a LoRA for the wrong base will load and do nothing. Keep a pool to one architecture family unless mixed results are the point of the experiment.