limbicnation
3bfc4f7db7
The post-execution cleanup unconditionally called unload_model(), which sends a generate request to evict the model. On the streaming-success path keep_alive="0s" has already evicted it, so that request forced Ollama to reload the model just to unload it again — a cold load+unload cycle with a VRAM spike after every node run. Track whether the subprocess fallback ran and pass unload=used_subprocess to cleanup. The streaming path relies on keep_alive="0s"; only the subprocess path (which carries the 5m default) still needs an explicit unload. CUDA cache release continues to run on every path.
ComfyUI Prompt Generator
Generate detailed Stable Diffusion prompts using Qwen3-8B via Ollama, directly within ComfyUI.
Features
- 7 Style Presets: Cinematic, Anime, Photorealistic, Fantasy, Abstract, Cyberpunk, Sci-Fi
- Temperature & Top-P Controls: Fine-tune generation creativity
- Focus Area (Emphasis): Direct the prompt to emphasize specific aspects
- Mood/Atmosphere: Set the emotional tone of the generated prompt
- Reasoning Toggle: Show or hide the model's thinking process
Prerequisites
-
Ollama installed and running: https://ollama.ai
-
Qwen3-8B model pulled:
ollama pull qwen3:8b
Installation
Option 1: Clone to custom_nodes (Recommended)
cd /path/to/ComfyUI/custom_nodes
git clone https://github.com/Limbicnation/ComfyUI-PromptGenerator.git
cd ComfyUI-PromptGenerator
pip install -r requirements.txt
Option 2: ComfyUI Manager
Search for "Prompt Generator" in the ComfyUI Manager and install.
Note: If you see "With the current security level configuration, only custom nodes from the 'default channel' can be installed", temporarily set security_level = weak in your ComfyUI Manager's config.ini file, then restore it to normal after installation. See PUBLISHING.md for details.
Usage
- Restart ComfyUI after installation
- Right-click → Add Node →
text/generation→ 🎨 Prompt Generator (Qwen) - Connect the output
promptto your text encoder or save node
Node Inputs
| Input | Type | Description |
|---|---|---|
description |
STRING | Your image concept (required) |
style |
COMBO | Style preset (cinematic, anime, etc.) |
emphasis |
STRING | Focus area (optional) |
mood |
STRING | Atmosphere/mood (optional) |
temperature |
FLOAT | Creativity (0.1-1.0, default: 0.7) |
top_p |
FLOAT | Sampling threshold (0.1-1.0, default: 0.9) |
include_reasoning |
BOOLEAN | Show model's thinking process |
model |
STRING | Ollama model (default: qwen3:8b) |
Example Workflow
Download the example workflow: workflow/workflow.png
Customizing Styles
Edit config/templates.yaml to add or modify style templates. Templates use Jinja2 syntax:
my_custom_style:
name: "My Style"
description: "Description for UI"
template: |
Write a prompt for: {{ description }}
{% if emphasis %}Focus on: {{ emphasis }}{% endif %}
{% if mood %}Mood: {{ mood }}{% endif %}
Troubleshooting
"Ollama not found"
- Ensure Ollama is installed and the
ollamacommand is in your PATH - Start the Ollama server:
ollama serve
"Model not found"
- Pull the model:
ollama pull qwen3:8b
Temperature/Top-P not working
- Install the Ollama Python package:
pip install ollama
License
Apache 2.0 - See LICENSE for details.
Related Projects
- Prompt Generator (Gradio App) - Standalone Gradio web UI for prompt generation
Credits
This ComfyUI node is based on the prompt-gen project.