Legion: Master (3 channels)
The tidy 3-channel Master for small worker jobs
- legion_config
- legion_campaign
- input_1
- input_2
- input_3
- legion_campaign
- output_1
- output_2
- output_3
LegionMaster3 is just Legion: Master with fewer ports. Where the 12-channel Master exposes input_1 … input_12, this variant gives you three inputs and three outputs plus the usual legion_campaign. Same execution logic, same config, same sync/async behavior - the only difference is a smaller face on the canvas.
Reach for it when your worker workflow only needs a couple of values. The classic case is a face-restore worker: LoadImage → Master3.input_1 and maybe a strength float on input_2. There's no performance difference between the variants - the extra ports on the 12-channel version are just wires you'd leave unconnected anyway. A smaller port surface means a cleaner graph and fewer chances to mis-wire something.
Inputs and outputs
legion_config(optional) - from Legion: Configuration. Either this orlegion_campaign, not both.legion_campaign(optional) - reuse a campaign/worker instead of a fresh config.input_1…input_3(optional) - data to ship to the worker.- Outputs:
legion_campaign, plusoutput_1…output_3- real results in sync mode, error-message strings in async mode (then collect via Legion: Join).
Same limits as the big Master: only images and primitives serialize, and the worker workflow must still start with Legion: Importer and end with Legion: Exporter.
Installing
Identical to the rest of the pack - ComfyUI Manager search "LegionPower", or:
cd ComfyUI/custom_nodes
git clone https://github.com/Transhumai/ComfyUI-LegionPower.git
cd ComfyUI-LegionPower
pip install -r requirements.txt
Restart ComfyUI. Just requests; nothing to download.
Troubleshooting
You inherit every pitfall of the full Master: Worker failed to start in time, Output manifest not found (no connected Exporter), and None results from unsupported data types. The fixes are the same too - check the console for the launch command, make sure the worker workflow is complete, and stick to images/primitives across the wire.
Pick 3 vs 6 vs 12 purely on how many values you actually move. Most face-restore and single-image post jobs never exceed three, so honestly, this is the variant I'd reach for first.
Inputs (5)
| Name | Type | Default | Description |
|---|---|---|---|
| legion_configopt | LEGION_CONFIG | — | |
| legion_campaignopt | LEGION_CAMPAIGN | — | |
| input_1opt | * | — | |
| input_2opt | * | — | |
| input_3opt | * | — |
Outputs (4)
| Name | Type | Description |
|---|---|---|
| legion_campaign | LEGION_CAMPAIGN | — |
| output_1 | * | — |
| output_2 | * | — |
| output_3 | * | — |