Prompt (VixUI)

If you're exporting ComfyUI workflows to Visionatrix, this is the node your flows live or die by. VixUiPrompt is the textarea: whatever string you put in it becomes the prompt box that a Visionatrix user sees and edits in the simplified UI, and the STRING output carries that text straight into your CLIPTextEncode. It's the difference between "the user can write their own prompt" and "the user is stuck with whatever you baked in."

The mechanism is deliberately boring - it passes text through unchanged - because the value is in the surrounding metadata. For anyone not targeting Visionatrix, this node is overkill: the stock CLIPTextEncode prompt box does the same job in plain ComfyUI.

Inputs that matter

• text (STRING, multiline) - the prompt itself. Note it's configured with dynamicPrompts enabled in the source, so {red|blue|green} wildcard syntax and __name__ wildcard files still work when the flow runs. That's a nice touch most flow authors don't expect.
• display_name (STRING, default "Prompt") - the label over the textarea in the Visionatrix UI.
• optional (BOOLEAN, default false) - unlike most VixUI controls this defaults to not optional, which makes sense: a flow without a prompt is a broken flow.
• advanced (BOOLEAN, default false) - keep it on the main form.
• order (INT, default 10) - placement; prompts usually belong near the top, and the default agrees.
• custom_id (STRING) - stable identifier for the control.

There are two optional inputs: hidden (invisible in the UI) and translatable - set translatable to true and Visionatrix can offer a translate button on the prompt box, which matters more than you'd think if your flows reach non-English users.

Outputs

A single STRING output. Wire it into the prompt input of your CLIPTextEncode - that's the entire intended usage.

How it works

do_it returns the text unchanged. All the work is upstream: Visionatrix scans the workflow, finds this node, and renders a textarea bound to the text input. The node exists so the flow author can declare the control in a way Visionatrix can find, instead of Visionatrix guessing which of a hundred string inputs is the prompt.

Install

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

Then restart ComfyUI - or grab ComfyUI-Visionatrix from ComfyUI Manager. No models to download; the dependency list is just torch, pillow and numpy, all already present.

Gotchas

Two things bite people. First, this node is the prompt source - don't also type a prompt into your CLIPTextEncode, or you'll have two competing texts and the upstream one wins (or worse, both). Second, if translatable is on but your backend isn't set up for translations, you may see an extra button that does nothing; leave it off unless you know your Visionatrix instance supports it. Otherwise: set display_name, wire the output to CLIPTextEncode, done.