LoRA Stack (LTX-2, 3 slots)

LTX-2 is fast - fast enough that you can actually iterate on it - which is exactly why you want a clean way to stack multiple LoRAs without a wall of chained LoraLoader nodes. IAMCCS_LTX2_LoRAStack is the LTX-2 sibling of the pack's WAN stack: three slots, each with its own strength, all bundled into one LORA output. The twist is that it runs every selected LoRA through LTX-2 key standardization before bundling, because LTX-2 LoRAs come in a couple of different key layouts and a mismatch means silent no-ops.

The inputs

Three slots, no model_type dial - LTX-2's key handling is handled automatically:

• lora1–lora3, with strength1–strength3. "no" leaves a slot empty; strength 0.0 skips the LoRA.
• Optional lora input - chain a previously built stack in for more than three LoRAs, same as the WAN stack's concatenation trick.

Output is a single LORA bundle. The node logs a friendly success line with how many LoRAs loaded and how many total parameter keys are active - handy when you're trying to figure out whether a LoRA actually got applied.

The mechanism is in standardize_ltx2_lora_keys: each .safetensors is loaded and its weights normalized to the key format the pack's LTX-2 apply node (IAMCCS_ModelWithLoRA_LTX2) expects. That standardization is the entire reason the stack exists - you pick LoRAs, it makes them compatible, and the apply node consumes the bundle.

Where it fits

LTX-2 workflows from the pack's 1.3.2 release use a stack + apply pair: build your bundle here, then feed it to Apply LoRA to MODEL (LTX-2, quiet logs) for the actual patching. If you never chain more than one LoRA, core's LoraLoader is fine and you don't need this. The moment you're stacking three style/character/quality LoRAs per run and swapping them between tests, the stack pays for itself - one node to edit instead of three chained loaders, plus the bundle stays reusable.

Install

Part of IAMCCS-nodes:

• ComfyUI Manager → search "IAMCCS" → install → restart.
• Or:

cd ComfyUI/custom_nodes
git clone https://github.com/IAMCCS/IAMCCS-nodes.git

Restart ComfyUI. Requirements: ComfyUI ≥ 0.3.0, Python ≥ 3.12, Torch ≥ 2.8, plus the upstream LTX-2 ComfyUI integration for the model itself. LoRAs come from your own loras folder; no downloads.