Merge LoRA Stacks

Merge LoRA Stacks

Back to Wiki Home

Overview

The Merge LoRA Stacks node merges multiple lora_stack inputs into a single lora_stack output. The left-side lora_stack inputs are dynamic (based on node API V3's Autogrow) and can be freely added or removed.

• Class name: zyd232 MergeLoraStacks
• Category: zyd232 Nodes
• Output: LORA_STACK

---

Inputs

Input	Type	Description
lora_stack_0	LORA_STACK	A lora_stack to merge (can be added dynamically)
lora_stack_1	LORA_STACK	A lora_stack to merge
...	LORA_STACK	More inputs, can be added dynamically

The left-side inputs are dynamic (Autogrow). Click the + on the node to add more input slots.

---

Output

Output	Type	Description
lora_stack	LORA_STACK	The merged single lora_stack

---

How It Works

• Merge order: Entries are concatenated in slot order (lora_stack_0 → lora_stack_1 → ...).
• Filtering: Entries with strength_model == 0 are automatically filtered out (supports both tuple (lora_name, strength_model, strength_clip) and dict formats).
• Empty inputs: None or empty lists are skipped and do not affect the result.

Data format

A lora_stack is a list of LoRA entries, each a 3-tuple:

(lora_name, strength_model, strength_clip)

For example:

[("lora_a.safetensors", 1.0, 1.0), ("lora_b.safetensors", 0.5, 0.5)]

---

Limits

• Maximum number of left-side lora_stack inputs: 100 (hard limit from ComfyUI V3 API's Autogrow._MaxNames = 100).
• Maximum total number of LoRAs after merging: no hard limit — limited only by memory (a lora_stack is a Python list concatenation).

---

Usage Examples

• Merge multiple LoRA stacks: connect the outputs of several LoRA Stacker nodes to this node's multiple inputs, then pass the merged stack to a downstream LoRA application node.
• Compatibility with legacy nodes: this node's LORA_STACK output uses the same type identifier as legacy V1 nodes' LORA_STACK, so it can be wired directly to/from nodes in comfyui-easy-use, comfyui-lora-optimizer, and others.

---

Back to Wiki Home