VNCCS LoRA Stack

Character LoRAs, style LoRAs, a pose LoRA if you're using Pose Studio's - a real VNCCS pipeline can end up stacking several LoRAs on top of one checkpoint, and chaining individual LoraLoader nodes for each one gets tedious fast. VNCCS LoRA Stack does what its name says: takes a model and clip, applies a whole list of LoRAs in one node, and hands back the patched pair.

What it does

Feed it your model and clip, give it a loras list, and it applies each one in order and returns the combined result. It's functionally the same idea as the popular general-purpose LoRA stack nodes you'll find in other packs (rgthree's Power Lora Loader is the best-known example across the ecosystem) - VNCCS ships its own so it can live inside its own category and stay consistent with the rest of the suite's widget-driven config style, rather than making you reach into a different pack mid-VNCCS-graph.

Inputs and outputs

• model (MODEL) and clip (CLIP) - the checkpoint pair to patch. Comes from wherever you're loading your base model - inside VNCCS's own workflows that's usually the model/clip Control Center exposes as optional inputs, if you're not letting Control Center manage the checkpoint itself.
• loras (STRING, default []) - the stack, managed through the node's widget rather than a raw text field. Add entries, set each LoRA's strength, toggle individual ones on or off without deleting them.
• Outputs: model and clip - the same pair, with every enabled LoRA in the stack applied. Wire these into your sampler and text encoder as you would any patched model/clip.

Installing it

Ships with the base pack. ComfyUI Manager: search "VNCCS - Visual Novel Character Creation Suite", install latest, restart. Manual: cd ComfyUI/custom_nodes && git clone https://github.com/AHEKOT/ComfyUI_VNCCS.git, restart. You'll need your actual LoRA files sitting in ComfyUI's standard models/loras folder as usual - this node stacks LoRAs, it doesn't fetch them.

Common issues & troubleshooting

Stacking more than two or three LoRAs degrades quality noticeably. This isn't a VNCCS bug, it's a property of the base model you're stacking onto. Anima in particular tolerates multiple LoRAs well as long as you don't max out every weight at once - keep individual strengths modest when several are active together rather than running each at full strength. Illustrious-family checkpoints are generally more forgiving of a heavier stack, but the same rule of thumb - don't run everything at 1.0 simultaneously - still applies.

A LoRA in the stack seems to do nothing. Check the entry is actually enabled (it's easy to toggle one off while testing and forget to re-enable it), and confirm the LoRA was actually trained for the checkpoint version you're running - a LoRA trained for an older Anima preview or an Illustrious v0.1-only style can behave inconsistently on a newer base.

Output looks fine with one LoRA but breaks with several. Test in isolation first: disable all but one, confirm it looks right, then add the rest back one at a time. It's the fastest way to find which specific LoRA in the stack is fighting the others rather than staring at the combined result guessing.