Pass Conditioning

Pass Conditioning is a pass-through node for the CONDITIONING data type - the encoded prompt tensor that comes out of a CLIP Text Encode node and goes into your sampler. It takes a conditioning in, hands you the exact same conditioning out, and does absolutely nothing in between. Which, as with the other passers in this pack, is precisely the point.

You reach for it in two situations. The first is when a workflow starts misbehaving after you bypass (Ctrl+B) a node. Bypassing is ComfyUI's cleanest way to A/B test - turn a region off without deleting it - but it has a nasty habit of breaking downstream connections to the bypassed node's inputs. Drop a passer in front of the fragile node and the conditioning still flows through a stable wire, so a bypass upstream stops being able to kill the graph.

The second is group hygiene. Managed groups need at least two nodes to exist, and they're a great way to tuck a pile of connections into one tidy box. A passer gives you a minimal second node to put in the group, lets you collapse it, and can fan a single conditioning out to multiple samplers - positive and negative from one text encoder, sent to a KSampler and an Advanced KSampler at once, without spaghetti.

Inputs and outputs

There's exactly one of each, which makes this node very hard to misuse:

• conditioning (CONDITIONING, required) - the encoded prompt. Wire it from the CONDITIONING output of a CLIP Text Encode (or from a merge/strength node).
• Output CONDITIONING - goes to the sampler's positive/negative input, or wherever the conditioning was headed.

Because it's pass-through, your values (text, conditioning strength, CFG) are untouched. Nothing to configure, nothing to tune.

Installing it

This node comes with ComfyUI-RvTools_v2, which installs as one pack:

cd ComfyUI/custom_nodes
git clone https://github.com/r-vage/ComfyUI-RvTools_v2

Restart ComfyUI after cloning, or just use ComfyUI Manager → search "RvTools" → Install → Restart. No model downloads involved. The pack's requirements (opencv-python, pilgram, pynvml, piexif on top of the usual torch/numpy/Pillow) are handled by Manager on install.

What to watch for

The README is upfront that this pack is no longer maintained, superseded by the same author's ComfyUI_Eclipse, which carries most of these nodes forward. The v2 pack still runs fine - you're just on the legacy train. If you're picking a starter pack today, Eclipse is the one to reach for.

The one real trap with passers: they add a hop to your graph, and if the node feeding them is bypassed, the passer passes through whatever it last got - which is usually fine, but don't expect it to conjure a value out of nowhere. And remember this is the CONDITIONING passer specifically. The pack has separate passers for CLIP, float, integer, string, latent, and so on. Grabbing the wrong one gives you a type mismatch and a red connection, which is at least a clear error rather than a silent one.