Shima SDXL Tuple Pass

The Shima SDXL Tuple Pass is a true pass-through: it takes an SDXL_TUPLE input and hands back the exact same tuple, unmodified. One wire in, one wire out, done. It's the reliable sibling of the pack's broken-looking Placeholder Tuple node - where that one fabricates placeholder junk, this one passes along the real thing, which makes it the one you actually want when you need the old Efficiency-Nodes-style tuple pipe in a Shima workflow.

SDXL_TUPLE is a legacy pipe type that bundles (model, clip, positive conditioning, negative conditioning) into a single wire - the pre-BNDL answer to the "twelve wires crossing the canvas" problem. Shima's own answer is the newer BNDL bundle system, but the pack keeps this compatibility node around so older workflows and third-party nodes that still speak SDXL_TUPLE keep working.

How it works

It's a few lines in nodes/legacy.py - the entire execute is:

def execute(self, **kwargs):
    return (kwargs.get("sdxl tuple", None),)

Both the input and the output are named sdxl tuple and typed SDXL_TUPLE. The input is optional, so an unwired input returns None rather than erroring. The tuple object itself is never inspected or modified - whatever Efficiency-Loader-style node produced it arrives at the other end bit-for-bit identical.

Where it fits

Three situations make it worth dropping in:

• Bridging legacy and modern nodes - decode a tuple from an older Efficient Loader and re-feed it after a detour through nodes that can't carry the tuple type.
• Island boundaries - like the pack's other passers, it's a stable landing pad for Use Everywhere broadcasts and a clear socket to aim wires at across a big graph.
• Graph organization - a named one-in/one-out node is a legible waypoint when a tuple wire has to cross the whole canvas.

If you're starting fresh and have no legacy tuple to carry, you don't need this node - the pack's BNDL packers (Model Citizen, Master Prompt, ReBNDLers) are the modern path.

Install

Same pack install as everything else:

cd ComfyUI/custom_nodes
git clone https://github.com/KDB-USJP/shima_wf.git Shima
pip install -r requirements.txt

Restart, find it under Shima/Utilities/Passers → Shima SDXL Tuple Pass. ComfyUI Manager: search "Shima". First boot auto-installs ComfyUI-Impact-Pack and cg-use-everywhere if missing - the island system's Use Everywhere dependency, not a problem.

Common issues

The classic confusion is mixing this up with Placeholder Tuple (same category, same SDXL_TUPLE output type). One passes real data; the other emits placeholder text with a tuple label. If a downstream node explodes with a type mismatch, check which one you grabbed. Also keep in mind it's a passive node: if the tuple upstream is stale (assembled before a LoRA was applied), this pass won't fix it - a stale tuple travels through a pass perfectly. The value you see on the output is whatever the input carried, which is the honest behavior you want from a pipe.