Pipe to Prompts

This is the unpacking half of the pair built around AV_PromptsToParametersPipe. That node collects your positive prompt, negative prompt, and optionally a reference image and mask into one PIPE bundle so they can travel through a big workflow as a single wire. This node is what opens the bundle back up: feed it a PIPE, get the individual pieces back out in the plain types your other nodes actually expect - text into a CLIP Text Encode, image and mask into whatever needs them.

It's plumbing, not a creative node in its own right, but it's the piece that makes the whole "carry everything as one wire" pattern actually usable. Without an unpacker, a PIPE full of prompt data is just an opaque object nothing downstream can read.

How it works

Give it a PIPE - typically the output of AV_PromptsToParametersPipe, possibly one that's already been extended with model selections from AV_CheckpointModelsToParametersPipe earlier in the chain - and it reads the prompt-related fields back out as their normal types. Nothing is transformed or interpreted; it's a straight read of whatever was packed in.

The inputs and outputs that matter

One required input:

• pipe - the PIPE to unpack.

Five outputs:

• pipe - the same bundle, passed straight through, so you can keep chaining more unpacking nodes (like AV_ParametersPipeToCheckpointModels for the model half) off the same wire instead of forking it.
• positive / negative - plain STRINGs, ready for a CLIP Text Encode.
• image - the reference IMAGE, if one was packed in.
• mask - the reference MASK, if one was packed in.

How to install it

Ships with the Art Venture pack. ComfyUI Manager: search comfyui-art-venture, install, restart. Manually:

cd ComfyUI/custom_nodes
git clone https://github.com/sipherxyz/comfyui-art-venture

Restart ComfyUI. No model downloads - this just reads a data structure.

Common issues & troubleshooting

positive/negative come out blank. The pipe you fed in was never built with AV_PromptsToParametersPipe (or whatever built it didn't set those fields), so there's nothing to unpack. Trace the pipe wire back to its source and confirm the prompt text actually went in somewhere upstream.

image/mask outputs are empty even though I attached one. Those two fields are optional on the builder side. If AV_PromptsToParametersPipe never had an image or mask wired into it, this node has nothing to hand back - it can't invent data that was never packed.

Using this node standalone with a hand-typed pipe. It's not meant to. This node only makes sense downstream of something that actually populated the bundle - pair it with AV_PromptsToParametersPipe (and, for the model side, AV_CheckpointModelsToParametersPipe / AV_ParametersPipeToCheckpointModels). On its own it just returns empty fields.

Feels like unnecessary indirection for a small workflow. It usually is, for a one-off. This pattern earns its keep in large, templated, or API-driven graphs where you want the entire "what to generate" manifest collapsed into one wire that an external caller can swap in one place. If your workflow is small, wiring a CLIP Text Encode directly is simpler.