26 lines
1.3 KiB
Markdown
26 lines
1.3 KiB
Markdown
# ComfyUI-LVLM-API
|
|
|
|
This workflow demonstrates a streamlined way to connect visual inputs with Large Multimodal Models (VLM) using a custom API node.
|
|
|
|
### **Workflow Breakdown**
|
|
|
|
* **Visual Source**: The **Load Image** node brings your local image into the ComfyUI environment.
|
|
* **The Prompt**: A **LLM (VLM) Prompt** node stores instructions. Here, it's configured as a "Visual Analysis Expert" to guide the AI on what details to focus on.
|
|
* **The Processing Hub**: The **LLM (VLM) API** node acts as the bridge. It leverages the **OpenAI API standard**, allowing you to connect to any compatible provider (e.g., OpenAI, Groq, DeepSeek, or local servers) by customizing the **base_url**.
|
|
* **The Output**: A **Show Text** node captures the model's response, displaying the generated description or analysis instantly.
|
|
|
|
### **In Action**
|
|
|
|
As seen in the screenshot, the system analyzes a portrait to describe the subject's attire (a grey puffer jacket), the indoor setting, and the soft lighting, turning a raw image into a detailed text report.
|
|
|
|
|
|
|
|
### **Installation & Usage**
|
|
|
|
To install this custom node, run the following commands in your terminal:
|
|
|
|
```bash
|
|
cd ComfyUI
|
|
git clone https://github.com/iwindy910/ComfyUI-LVLM-API.git
|
|
.\python_embeded\python.exe -m pip install -r custom_nodes\ComfyUI-LVLM-API\requirements.txt
|