Lora Cycler (LoraManager)
Step through LoRAs one at a time across a batch
- cycler_config
- pool_config
- LORA_STACK
Lora Cycler is the Randomizer's disciplined sibling. Where the Randomizer picks at random, the Cycler steps through your LoRAs in order - one (or a defined set) per run - so across a batch or a queue you get systematic coverage instead of a lucky-dip. If you want to see the same prompt through each of your character LoRAs, or sweep a folder of styles one by one, this is the node that does it without you manually swapping files between runs.
That's the difference that matters: reproducible, complete coverage. A randomizer might hit the same combo twice and never touch a third; the cycler marches through the list. It's the closest thing the pack has to an XY-plot-style sweep for LoRA selection, and it outputs a standard LORA_STACK so any loader can apply what it picked.
This is the node I'd reach for when I'm evaluating a batch of LoRAs I just downloaded and want an apples-to-apples comparison - same prompt, same seed, same everything, only the LoRA changing. Random selection is fun for discovery but useless for a fair test, because you can't control what it lands on. The Cycler gives you the controlled sweep.
How it works
You configure a cycler_config that defines the cycling behavior - the order and how it advances between runs - and optionally feed it a Lora Pool to cycle through. On each execution it advances to the next selection and emits that as a LORA_STACK. Chain that into a loader, queue up a batch, and each image comes out with the next LoRA in the sequence.
Inputs and outputs that matter
cycler_config(CYCLER_CONFIG, required) - the node's configuration widget defining what to cycle through and how it steps.pool_config(POOL_CONFIG, optional) - connect a Lora Pool to cycle across a reusable predefined pool.
One output: LORA_STACK - into any stack-consuming loader, or into Lora Stack Combiner to pair with fixed LoRAs.
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
It's not advancing between runs. The cycler steps on execution - make sure you're actually queuing new runs (a batch or repeated prompts), not re-rendering a cached result. If nothing downstream changed, ComfyUI may be serving a cached output; nudge the graph so it re-executes.
I want random, not sequential. Then you want Lora Randomizer. Same output type, opposite selection strategy - the Cycler is for orderly coverage.
A cycled LoRA does nothing. Same architecture rule as everywhere in this pack: if the sequence lands on a LoRA that doesn't match your checkpoint's base model, it loads and has no effect. Keep the cycled set to one family.
Inputs (2)
| Name | Type | Default | Description |
|---|---|---|---|
| cycler_config | CYCLER_CONFIG | — | |
| pool_configopt | POOL_CONFIG | — |
Outputs (1)
| Name | Type | Description |
|---|---|---|
| LORA_STACK | LORA_STACK | — |