PS Text Receiver (LLM)

PSTextReceiver is the smallest, most forgettable node in the ComfyUI-PS-Connector pack, and that's exactly the point. It's an output node built for one scenario: an LLM workflow that ends with text instead of an image, where that text needs to travel back into Photoshop. The pack's signature move - double-click a layer in PS and a vision model reverse-tags it into a prompt - ends at this node. It's the "delivery room" for the words.

What it does is almost insultingly simple. It takes a single required text input (a STRING, forced to be wired in from another node rather than typed by hand), and on execution it just hands the text back to ComfyUI's UI bus as {"ui": {"text": [text]}}. The PS frontend picks that up over the WebSocket and does whatever it does with it - in the default-mode flow, that's pasting a generated caption where you can read or reuse it. No outputs, an output node by declaration, and the README describes it as "connect at the final output, though generally not needed."

"Generally not needed" is doing a lot of work there, and it's worth understanding why before you add it everywhere. In the pack's (LLM)Qwen3.json example, the workflow is an image-to-text pipeline: a LoadImage feeding an AILab_QwenVL node, with PSTextReceiver at the end. The (LLM) prefix on the workflow filename is what tells the PS plugin to treat it as a captioning workflow in the first place. So this node isn't something you sprinkle into a normal txt2img graph - it belongs at the tail of a dedicated (LLM)-named workflow, and only if you actually want the text result surfaced in PS.

Why would you? The reverse-tagging flow. You're editing in PS, you double-click a layer, and instead of alt-tabbing to a web UI to generate tags for a LoRA you're about to train, the caption lands back in your panel. If you're running a pure ComfyUI-side workflow and just want to see the text in the console or a preview node, you don't need this - any Show Text-style node does that job. The only time it matters is when the PS connector is the consumer.

The input is the whole interface:

• text (STRING, required) - the string to hand back to the PS frontend. Wire it from the output of your LLM/vision node; it's set to forceInput so you can't (and shouldn't) type a constant in here.

Installation is the pack install, one time, shared by all six nodes: search "ComfyUI PS Connector" in ComfyUI Manager, or

cd ComfyUI/custom_nodes
git clone https://github.com/doubley2000/ComfyUI-PS-Connector

then restart ComfyUI. The plugin auto-installs ComfyUI-Lora-Manager and copies example workflows to ComfyUI/user/default/workflows/PSflows/ on first launch, so give it one more restart after that. No extra Python deps - the pack's pyproject.toml is empty on dependencies.

Troubleshooting is minimal because there's so little to break. If PS never shows the text, the usual culprits are: the workflow isn't named with a ( prefix (so the plugin never lists it), it isn't exported in API format, or the PS panel isn't pointed at your ComfyUI URL (http://127.0.0.1:8188 by default). This is a young pack from a solo, non-programmer author ("all code is AI-generated" is right there in the README), so when in doubt, treat the example (LLM)Qwen3.json workflow as your ground truth and copy its wiring.