🔥Gradio Negative Prompt
The negative prompt box for your packaged ComfyUI workflow
- STRING
GradioTextBad (🔥Gradio Negative Prompt) is the negative-prompt half of the ComfyUI_to_webui frontend. Where its sibling GradioTextOk can appear multiple times to spawn several positive-prompt boxes, this one is deliberately a singleton: exactly one negative prompt field in your Gradio UI, always. If you're packaging a workflow for someone else to run, that's actually the right design - people typo their negatives, and one stable field beats five.
The mechanism is a straight passthrough, same as its sibling. string in, string out, no encoding performed despite the "encode" function name. The text comes back as a STRING output that you wire into a normal CLIPTextEncode's text slot - the actual embedding happens there, like it always did. The name input just labels the node so the frontend and you can tell it apart; it doesn't change behavior.
Inputs, for a beginner:
string- the negative text. Multiline, supports dynamic prompts ({wildcards}etc.) if you've got that enabled.name- label for the frontend, defaults to "GradioTextBad".
Output: STRING, straight into CLIPTextEncode. In a typical packaged workflow the trio is 💧Gradio Positive Prompt → CLIPTextEncode, 🔥Gradio Negative Prompt → CLIPTextEncode, both feeding a KSampler.
Install is the pack install - ComfyUI Manager → "ComfyUI_to_webui", or clone into custom_nodes:
cd ComfyUI/custom_nodes
git clone https://github.com/kungful/ComfyUI_to_webui.git
then restart. The pack grabs gradio, opencv, Pillow, numpy and httpx on import (or via requirements.txt if the auto-install fails), and ships no models. Run the workflow once inside ComfyUI so the API JSON gets auto-saved to output/; the frontend then shows the negative field and feeds what you type back into this node on each run.
One honest caveat: because the frontend is built around exact node matching, this node is specifically for the pack's own Gradio UI. If you just want a plain text node in a normal ComfyUI graph, the stock CLIPTextEncode is already there and this adds nothing. Reach for it only when the goal is "give my workflow a clean webui front." It's the least interesting node in the pack to read about, and the most important one to include when the person on the other end of the UI isn't a ComfyUI user.
Inputs (2)
| Name | Type | Default | Description |
|---|---|---|---|
| string | STRING | The text to be encoded. | |
| name | STRING | GradioTextBad | 节点名称 |
Outputs (1)
| Name | Type | Description |
|---|---|---|
| STRING | STRING | A conditioning containing the embedded text used to guide the diffusion model. |