QWEN Prompter (Creepybits)

QWEN Prompter is meant to be the Qwen-specific member of the pack's "Creative Council" - a node that wraps your raw idea in instructions so a downstream LLM writes prompts in the style and syntax Qwen-Image handles best. The Qwen family is its own beast with its own prompt expectations, and this node is supposed to be the bridge.

It has one input (text_2), one output (text), and the mechanic is identical to Master Key and Keyword Extractor: load a fixed instruction file from assets/prompts, prepend it to your text, pass it through. The design is sound.

Here's the catch, and it's a real one: the file this node expects doesn't ship with the pack. The code loads QWEN_prompt.md from assets/prompts, but as of the current repo state that folder contains keywords.txt, prompt.txt, system_prompt.txt - no QWEN_prompt.md. So on a stock install, this node loads nothing, prints an error to the console, and returns ERROR: Prompt file not found. glued onto your text. I verified this against the live repository, not just the docs: the file is referenced by the code and absent from the assets.

How it works (and how it currently doesn't)

At startup it tries to read QWEN_prompt.md. On failure - which is the default state today - fixed_text becomes the literal string ERROR: Prompt file not found. Every run then concatenates that error text with whatever you typed into text_2. So the node "works" in the sense that it produces output; it just produces wrong output.

The fix is simple and makes the node genuinely useful: create assets/prompts/QWEN_prompt.md and write (or paste) the Qwen prompt-formatting instructions you want. Model it on the sibling files - prompt.txt in the same folder is a good template for tone and structure. Once the file exists, the node behaves exactly as designed: your idea in, Qwen-ready brief out, and the text output feeds the system prompt of your Gemini node, with your image on Gemini's image input if you want Qwen to work from a reference.

Why you'd bother

If you're building Qwen-Image workflows, prompt format genuinely matters - the Qwen-Image model expects structured, detailed descriptions, and this node exists to make sure every prompt you generate follows that house style. Paired with the pack's Qwen Aspect Ratio node (which handles the resolution half of Qwen-Image's requirements), you have the two halves of a Qwen-friendly graph.

Installing it

Part of ComfyUI-Creepy_nodes:

• ComfyUI Manager: search "Creepy_nodes" and install.
• Manual:

  cd ComfyUI/custom_nodes
  git clone https://github.com/Creepybits/ComfyUI-Creepy_nodes.git
  

Restart ComfyUI. No dependencies beyond what the Gemini node needs.

Troubleshooting

If your output starts with ERROR: Prompt file not found., you've hit the missing-file issue above - create QWEN_prompt.md in assets/prompts and the error disappears on the next run. If the error persists after creating it, check the filename is exact (case matters) and restart ComfyUI so the node re-reads it at init. And remember: no Gemini node downstream means no actual Qwen-style prompt - this node only shapes the instructions, the LLM does the writing.