@lobehub/chat/docs/usage/providers/comfyui.mdx

Lobe Chat - an open-source, high-performance chatbot framework that supports speech synthesis, multimodal, and extensible Function Call plugin system. Supports one-click free deployment of your private ChatGPT/LLM web application.

---
title: Using ComfyUI for Image Generation in LobeChat
description: Learn how to configure and use ComfyUI service in LobeChat, supporting FLUX series models for high-quality image generation and editing features
tags:
  - ComfyUI
  - FLUX
  - Text-to-Image
  - Image Editing
  - AI Image Generation
---

# Using ComfyUI in LobeChat

<Image alt={'Using ComfyUI in LobeChat'} cover src={'https://hub-apac-1.lobeobjects.space/docs/e9b811f248a1db2bd1be1af888cf9b9d.png'} />

This documentation will guide you on how to use [ComfyUI](https://github.com/comfyanonymous/ComfyUI) in LobeChat for high-quality AI image generation and editing.

## ComfyUI Overview

ComfyUI is a powerful stable diffusion and flow diffusion GUI that provides a node-based workflow interface. LobeChat integrates with ComfyUI, supporting complete FLUX series models, including text-to-image generation and image editing capabilities.

### Key Features

- **Extensive Model Support**: Supports 223 models, including FLUX series (130) and SD series (93)
- **Configuration-Driven Architecture**: Registry system provides intelligent model selection
- **Multi-Format Support**: Supports .safetensors and .gguf formats with various quantization levels
- **Dynamic Precision Selection**: Supports default, fp8\_e4m3fn, fp8\_e5m2, fp8\_e4m3fn\_fast precision
- **Multiple Authentication Methods**: Supports no authentication, basic authentication, Bearer Token, and custom authentication
- **Intelligent Component Selection**: Automatically selects optimal T5, CLIP, VAE encoder combinations
- **Enterprise-Grade Optimization**: Includes NF4, SVDQuant, TorchAO, MFLUX optimization variants

## Quick Start

### Step 1: Configure ComfyUI in LobeChat

#### 1. Open Settings Interface

- Access LobeChat's `Settings` interface
- Find the `ComfyUI` setting item under `AI Providers`

<Image alt={'ComfyUI Settings Interface'} inStep src={'https://github.com/lobehub/lobe-chat/assets/17870709/3f31bc33-509f-4ad2-ba81-280c2a6ec5fa'} />

#### 2. Configure Connection Parameters

**Basic Configuration**:

- **Server Address**: Enter ComfyUI server address, e.g., `http://localhost:8188`
- **Authentication Type**: Select appropriate authentication method (default: no authentication)

### Step 2: Select Model and Start Generating Images

#### 1. Select FLUX Model

In the conversation interface:

- Click the model selection button
- Select the desired FLUX model from the ComfyUI category

<Image alt={'Select FLUX Model'} inStep src={'https://github.com/lobehub/lobe-chat/assets/17870709/ff7ebacf-27f0-42d7-810b-00314499a084'} />

#### 2. Text-to-Image Generation

**Using FLUX Schnell (Fast Generation)**:

```plaintext
Generate an image: A cute orange cat sitting on a sunny windowsill, warm lighting, detailed fur texture
```

**Using FLUX Dev (High Quality Generation)**:

```plaintext
Generate high quality image: City skyline at sunset, cyberpunk style, neon lights, 4K high resolution, detailed architecture
```

#### 3. Image Editing

**Using FLUX Kontext-dev for Image Editing**:

```plaintext
Edit this image: Change the background to a starry night sky, keep the main subject, cosmic atmosphere
```

Then upload the original image you want to edit.

<Callout type={'info'}>
  Image editing functionality requires uploading the original image first, then describing the modifications you want to make.
</Callout>

## Authentication Configuration Guide

ComfyUI supports four authentication methods. Choose the appropriate method based on your server configuration and security requirements:

### No Authentication (none)

**Use Cases**:

- Local development environment (localhost)
- Internal network with trusted users
- Personal single-machine deployment

**Configuration**:

```yaml
Authentication Type: None
Server Address: http://localhost:8188
```

### Basic Authentication (basic)

**Use Cases**:

- Deployments using Nginx reverse proxy
- Team internal use requiring basic access control

**Configuration**:

1. **Create User Password**:

```bash
# Install apache2-utils
sudo apt-get install apache2-utils

# Create user 'admin'
sudo htpasswd -c /etc/nginx/.htpasswd admin
```

2. **LobeChat Configuration**:

```yaml
Authentication Type: Basic Authentication
Server Address: http://your-domain.com
Username: admin
Password: your_secure_password
```

### Bearer Token (bearer)

**Use Cases**:

- API-driven application integration
- Enterprise environments requiring Token authentication

**Generate Token**:

```python
import jwt
import datetime

payload = {
    'user': 'admin',
    'exp': datetime.datetime.utcnow() + datetime.timedelta(days=30)
}

secret_key = "your-secret-key"
token = jwt.encode(payload, secret_key, algorithm='HS256')
print(f"Bearer Token: {token}")
```

**LobeChat Configuration**:

```yaml
Authentication Type: Bearer Token
Server Address: http://your-server:8188
API Key: eyJhbGciOiJIUzI1NiIsInR5cCI6IkpXVCJ9...
```

### Custom Authentication (custom)

**Use Cases**:

- Integration with existing enterprise authentication systems
- Systems requiring multiple authentication headers

**LobeChat Configuration**:

```yaml
Authentication Type: Custom
Server Address: http://your-server:8188
Custom Headers:
{
  "X-API-Key": "your_api_key",
  "X-Client-ID": "lobechat"
}
```

## Common Issues Resolution

### 1. How to Install Comfy-Manager

Comfy-Manager is ComfyUI's extension manager that allows you to easily install and manage various nodes, models, and extensions.

<details>
  <summary><b>📦 Install Comfy-Manager Steps</b></summary>

  #### Method 1: Manual Installation (Recommended)

  ```bash
  # Navigate to ComfyUI's custom_nodes directory
  cd ComfyUI/custom_nodes

  # Clone Comfy-Manager repository
  git clone https://github.com/ltdrdata/ComfyUI-Manager.git

  # Restart ComfyUI server
  # After restart, you'll see the Manager button in the UI
  ```

  #### Method 2: One-Click Installation Script

  ```bash
  # Execute in ComfyUI root directory
  curl -fsSL https://raw.githubusercontent.com/ltdrdata/ComfyUI-Manager/main/install.sh | bash
  ```

  #### Verify Installation

  1. Restart ComfyUI server
  2. Visit `http://localhost:8188`
  3. You should see the "Manager" button in the bottom-right corner

  #### Using Comfy-Manager

  **Install Models**:

  1. Click "Manager" button
  2. Select "Install Models"
  3. Search for needed models (e.g., FLUX, SD3.5)
  4. Click "Install" to automatically download to correct directory

  **Install Node Extensions**:

  1. Click "Manager" button
  2. Select "Install Custom Nodes"
  3. Search for needed nodes (e.g., ControlNet, AnimateDiff)
  4. Click "Install" and restart server

  **Manage Installed Content**:

  1. Click "Manager" button
  2. Select "Installed" to view installed extensions
  3. Update, disable, or uninstall extensions
</details>

### 2. How to Handle "Model not found" Errors

When you see errors like `Model not found: flux1-dev.safetensors, flux1-krea-dev.safetensors, flux1-schnell.safetensors`, it means the required model files are missing from the server.

<details>
  <summary><b>🔧 Resolve Model not found Errors</b></summary>

  #### Error Example

  ```plaintext
  Model not found: flux1-dev.safetensors, flux1-krea-dev.safetensors, flux1-schnell.safetensors
  ```

  This error indicates the system expects to find these model files but couldn't locate them on the server.

  #### Resolution Methods

  **Method 1: Download using Comfy-Manager (Recommended)**

  1. Open ComfyUI interface
  2. Click "Manager" → "Install Models"
  3. Search for the model name from the error (e.g., "flux1-dev")
  4. Click "Install" to automatically download

  **Method 2: Manual Model Download**

  1. **Download Model Files**:
     - Visit [Hugging Face](https://huggingface.co/black-forest-labs/FLUX.1-dev) or other model sources
     - Download the files mentioned in the error (e.g., `flux1-dev.safetensors`)

  2. **Place in Correct Directory**:
     ```bash
     # FLUX and SD3.5 main models go to
     ComfyUI/models/diffusion_models/flux1-dev.safetensors

     # SD1.5 and SDXL models go to
     ComfyUI/models/checkpoints/
     ```

  3. **Verify Files**:
     ```bash
     # Check if file exists
     ls -la ComfyUI/models/diffusion_models/flux1-dev.safetensors

     # Check file integrity (optional)
     sha256sum flux1-dev.safetensors
     ```

  4. **Restart ComfyUI Server**

  **Method 3: Direct Download with wget/curl**

  ```bash
  # Navigate to models directory
  cd ComfyUI/models/diffusion_models/

  # Download using wget (replace with actual download link)
  wget https://huggingface.co/black-forest-labs/FLUX.1-dev/resolve/main/flux1-dev.safetensors

  # Or use curl
  curl -L -o flux1-dev.safetensors https://huggingface.co/black-forest-labs/FLUX.1-dev/resolve/main/flux1-dev.safetensors
  ```

  #### Common Model Download Sources

  - **Hugging Face**: [https://huggingface.co/models](https://huggingface.co/models)
  - **Civitai**: [https://civitai.com/models](https://civitai.com/models)
  - **Official Sources**:
    - FLUX: [https://huggingface.co/black-forest-labs](https://huggingface.co/black-forest-labs)
    - SD3.5: [https://huggingface.co/stabilityai](https://huggingface.co/stabilityai)

  #### Prevention Measures

  1. **Basic Model Package**: Download at least one base model
     - FLUX: `flux1-schnell.safetensors` (fast) or `flux1-dev.safetensors` (high quality)
     - SD3.5: `sd3.5_large.safetensors`

  2. **Check Disk Space**:
     ```bash
     # Check available space
     df -h ComfyUI/models/
     ```

  3. **Set Model Path** (optional):
     If your models are stored elsewhere, create symbolic links:
     ```bash
     ln -s /path/to/your/models ComfyUI/models/diffusion_models/
     ```
</details>

### 3. How to Handle Missing System Component Errors

When you see errors like `Missing VAE encoder: ae.safetensors` or other component files missing, you need to download the corresponding system components.

<details>
  <summary><b>🛠️ Resolve Missing System Component Errors</b></summary>

  #### Common Component Errors

  ```plaintext
  Missing VAE encoder: ae.safetensors. Please download and place it in the models/vae folder.
  Missing CLIP encoder: clip_l.safetensors. Please download and place it in the models/clip folder.
  Missing T5 encoder: t5xxl_fp16.safetensors. Please download and place it in the models/clip folder.
  ```

  #### Component Types Description

  | Component Type | Example Filename               | Purpose                 | Storage Directory  |
  | -------------- | ------------------------------ | ----------------------- | ------------------ |
  | **VAE**        | ae.safetensors                 | Image encoding/decoding | models/vae/        |
  | **CLIP**       | clip\_l.safetensors            | Text encoding (CLIP)    | models/clip/       |
  | **T5**         | t5xxl\_fp16.safetensors        | Text encoding (T5)      | models/clip/       |
  | **ControlNet** | flux-controlnet-\*.safetensors | Control networks        | models/controlnet/ |

  #### Resolution Methods

  **Method 1: Use Comfy-Manager (Recommended)**

  1. Click "Manager" → "Install Models"
  2. Select component type in "Filter" (VAE/CLIP/T5)
  3. Download corresponding component files

  **Method 2: Manual Component Download**

  ##### FLUX Required Components

  ```bash
  # 1. VAE Encoder
  cd ComfyUI/models/vae/
  wget https://huggingface.co/black-forest-labs/FLUX.1-dev/resolve/main/ae.safetensors

  # 2. CLIP-L Encoder
  cd ComfyUI/models/clip/
  wget https://huggingface.co/comfyanonymous/flux_text_encoders/resolve/main/clip_l.safetensors

  # 3. T5-XXL Encoder (choose different precisions)
  # FP16 version (recommended, balanced performance)
  wget https://huggingface.co/comfyanonymous/flux_text_encoders/resolve/main/t5xxl_fp16.safetensors

  # Or FP8 version (saves VRAM)
  wget https://huggingface.co/comfyanonymous/flux_text_encoders/resolve/main/t5xxl_fp8_e4m3fn.safetensors
  ```

  ##### SD3.5 Required Components

  ```bash
  # SD3.5 uses different encoders
  cd ComfyUI/models/clip/

  # CLIP-G Encoder
  wget https://huggingface.co/stabilityai/stable-diffusion-3.5-large/resolve/main/text_encoders/clip_g.safetensors

  # CLIP-L Encoder
  wget https://huggingface.co/stabilityai/stable-diffusion-3.5-large/resolve/main/text_encoders/clip_l.safetensors

  # T5-XXL Encoder
  wget https://huggingface.co/stabilityai/stable-diffusion-3.5-large/resolve/main/text_encoders/t5xxl_fp16.safetensors
  ```

  ##### SDXL Required Components

  ```bash
  # SDXL VAE
  cd ComfyUI/models/vae/
  wget https://huggingface.co/stabilityai/sdxl-vae/resolve/main/sdxl_vae.safetensors

  # SDXL uses built-in CLIP encoders, usually no separate download needed
  ```

  #### Component Compatibility Matrix

  | Model Series | Required VAE   | Required CLIP       | Required T5             | Optional Components |
  | ------------ | -------------- | ------------------- | ----------------------- | ------------------- |
  | **FLUX**     | ae.safetensors | clip\_l.safetensors | t5xxl\_fp16.safetensors | ControlNet          |
  | **SD3.5**    | Built-in       | clip\_g + clip\_l   | t5xxl\_fp16             | -                   |
  | **SDXL**     | sdxl\_vae      | Built-in            | -                       | Refiner             |
  | **SD1.5**    | vae-ft-mse     | Built-in            | -                       | ControlNet          |

  #### Precision Selection Recommendations

  **T5 Encoder Precision Selection**:

  | VRAM Capacity | Recommended Version | Filename                       |
  | ------------- | ------------------- | ------------------------------ |
  | \< 12GB       | FP8 Quantized       | t5xxl\_fp8\_e4m3fn.safetensors |
  | 12-16GB       | FP16                | t5xxl\_fp16.safetensors        |
  | > 16GB        | FP32                | t5xxl.safetensors              |

  #### Verify Component Installation

  ```bash
  # Check all required components
  echo "=== VAE Components ==="
  ls -la ComfyUI/models/vae/

  echo "=== CLIP/T5 Components ==="
  ls -la ComfyUI/models/clip/

  echo "=== ControlNet Components ==="
  ls -la ComfyUI/models/controlnet/
  ```

  #### Troubleshooting

  **Issue: Still getting errors after download**

  1. **Check File Permissions**:
     ```bash
     chmod 644 ComfyUI/models/vae/*.safetensors
     chmod 644 ComfyUI/models/clip/*.safetensors
     ```

  2. **Clear Cache**:
     ```bash
     # Clear ComfyUI cache
     rm -rf ComfyUI/temp/*
     rm -rf ComfyUI/__pycache__/*
     ```

  3. **Restart Server**:
     ```bash
     # Fully restart ComfyUI
     pkill -f "python main.py"
     python main.py --listen 0.0.0.0 --port 8188
     ```

  **Issue: Insufficient VRAM**

  Use quantized component versions:

  - T5: Use `t5xxl_fp8_e4m3fn.safetensors` instead of FP16/FP32
  - VAE: Some models support FP16 VAE versions

  **Issue: Slow Downloads**

  1. Use mirror sources (if applicable)
  2. Use download tools (like aria2c) with resume support:
     ```bash
     aria2c -x 16 -s 16 -k 1M [download_link]
     ```
</details>

## ComfyUI Server Installation

<details>
  <summary><b>🚀 Install and Configure ComfyUI Server</b></summary>

  ### 1. Install ComfyUI

  ```bash
  # Clone ComfyUI repository
  git clone https://github.com/comfyanonymous/ComfyUI.git
  cd ComfyUI

  # Install dependencies
  pip install -r requirements.txt

  # Optional: Install JWT support (for Token authentication)
  pip install PyJWT

  # Start ComfyUI server
  python main.py --listen 0.0.0.0 --port 8188
  ```

  ### 2. Download Model Files

  **Recommended Basic Configuration** (Minimal installation):

  **Main Models** (place in `models/diffusion_models/` directory):

  - `flux1-schnell.safetensors` - Fast generation (4 steps)
  - `flux1-dev.safetensors` - High-quality creation (20 steps)

  **Required Components** (place in respective directories):

  - `models/vae/ae.safetensors` - VAE encoder
  - `models/clip/clip_l.safetensors` - CLIP text encoder
  - `models/clip/t5xxl_fp16.safetensors` - T5 text encoder

  ### 3. Verify Server Running

  Visit `http://localhost:8188` to confirm ComfyUI interface loads properly.

  <Callout type={'info'}>
    **Smart Model Selection**: LobeChat will automatically select the best model based on available model files on the server. You don't need to download all models; the system will automatically choose from available models by priority (Official > Enterprise > Community).
  </Callout>
</details>

## Supported Models

LobeChat's ComfyUI integration uses a configuration-driven architecture, supporting **223 models**, providing complete coverage from official models to community-optimized versions.

### FLUX Series Recommended Parameters

| Model Type  | Recommended Steps | CFG Scale | Resolution Range     |
| ----------- | ----------------- | --------- | -------------------- |
| **Schnell** | 4 steps           | -         | 512×512 to 1536×1536 |
| **Dev**     | 20 steps          | 3.5       | 512×512 to 2048×2048 |
| **Kontext** | 20 steps          | 3.5       | 512×512 to 2048×2048 |
| **Krea**    | 20 steps          | 4.5       | 512×512 to 2048×2048 |

### SD3.5 Series Parameters

| Model Type      | Recommended Steps | CFG Scale | Resolution Range     |
| --------------- | ----------------- | --------- | -------------------- |
| **Large**       | 25 steps          | 7.0       | 512×512 to 2048×2048 |
| **Large Turbo** | 8 steps           | 3.5       | 512×512 to 1536×1536 |
| **Medium**      | 20 steps          | 6.0       | 512×512 to 1536×1536 |

<details>
  <summary><b>📋 Complete Supported Model List</b></summary>

  ### Model Classification System

  #### Priority 1: Official Core Models

  **FLUX.1 Official Series**:

  - `flux1-dev.safetensors` - High-quality creation model
  - `flux1-schnell.safetensors` - Fast generation model
  - `flux1-kontext-dev.safetensors` - Image editing model
  - `flux1-krea-dev.safetensors` - Safety-enhanced model

  **SD3.5 Official Series**:

  - `sd3.5_large.safetensors` - SD3.5 large base model
  - `sd3.5_large_turbo.safetensors` - Fast generation version
  - `sd3.5_medium.safetensors` - Medium-scale model

  #### Priority 2: Enterprise Optimized Models (106 FLUX)

  **Quantization Optimization Series**:

  - **GGUF Quantization**: Each variant supports 11 quantization levels (F16, Q8\_0, Q6\_K, Q5\_K\_M, Q5\_K\_S, Q4\_K\_M, Q4\_K\_S, Q4\_0, Q3\_K\_M, Q3\_K\_S, Q2\_K)
  - **FP8 Precision**: fp8\_e4m3fn, fp8\_e5m2 optimized versions
  - **Enterprise Lightweight**: FLUX.1-lite-8B series
  - **Technical Experiments**: NF4, SVDQuant, TorchAO, optimum-quanto, MFLUX optimized versions

  #### Priority 3: Community Fine-tuned Models (48 FLUX)

  **Community Optimization Series**:

  - **Jib Mix Flux** Series: High-quality mixed models
  - **Real Dream FLUX** Series: Realism style
  - **Vision Realistic** Series: Visual realism
  - **PixelWave FLUX** Series: Pixel art optimization
  - **Fluxmania** Series: Diverse style support

  ### SD Series Model Support (93 models)

  **SD3.5 Series**: 5 models
  **SD1.5 Series**: 37 models (including official, quantized, and community versions)
  **SDXL Series**: 50 models (including base, Refiner, and Playground models)

  ### Workflow Support

  System supports **6 workflows**:

  - **flux-dev**: High-quality creation workflow
  - **flux-schnell**: Fast generation workflow
  - **flux-kontext**: Image editing workflow
  - **sd35**: SD3.5 dedicated workflow
  - **simple-sd**: Simple SD workflow
  - **index**: Workflow entry point
</details>

## Performance Optimization Recommendations

### Hardware Requirements

**Minimum Configuration** (GGUF quantized models):

- GPU: 6GB VRAM (using Q4 quantization)
- RAM: 12GB
- Storage: 30GB available space

**Recommended Configuration** (standard models):

- GPU: 12GB+ VRAM (RTX 4070 Ti or higher)
- RAM: 24GB+
- Storage: SSD 100GB+ available space

### VRAM Optimization Strategy

| VRAM Capacity | Recommended Quantization | Model Example                      | Performance Characteristics |
| ------------- | ------------------------ | ---------------------------------- | --------------------------- |
| **6-8GB**     | Q4\_0, Q4\_K\_S          | `flux1-dev-Q4_0.gguf`              | Minimal VRAM usage          |
| **10-12GB**   | Q6\_K, Q8\_0             | `flux1-dev-Q6_K.gguf`              | Balance performance/quality |
| **16GB+**     | FP8, FP16                | `flux1-dev-fp8-e4m3fn.safetensors` | Near-original quality       |
| **24GB+**     | Full model               | `flux1-dev.safetensors`            | Best quality                |

## Custom Model Usage

<details>
  <summary><b>🎨 Configure Custom SD Models</b></summary>

  LobeChat supports using custom Stable Diffusion models. The system uses fixed filenames to identify custom models.

  ### 1. Model File Preparation

  **Required Files**:

  - **Main Model File**: `custom_sd_lobe.safetensors`
  - **VAE File (Optional)**: `custom_sd_vae_lobe.safetensors`

  ### 2. Add Custom Model

  **Method 1: Rename Existing Model**

  ```bash
  # Rename your model to fixed filename
  mv your_custom_model.safetensors custom_sd_lobe.safetensors

  # Move to correct directory
  mv custom_sd_lobe.safetensors ComfyUI/models/diffusion_models/
  ```

  **Method 2: Create Symbolic Link (Recommended)**

  ```bash
  # Create soft link for easy model switching
  ln -s /path/to/your_model.safetensors ComfyUI/models/diffusion_models/custom_sd_lobe.safetensors
  ```

  ### 3. Use Custom Model

  In LobeChat, custom models will appear as:

  - **stable-diffusion-custom**: Standard custom model
  - **stable-diffusion-custom-refiner**: Refiner custom model

  ### Custom Model Parameter Recommendations

  | Parameter  | SD 1.5 Models | SDXL Models |
  | ---------- | ------------- | ----------- |
  | **steps**  | 20-30         | 25-40       |
  | **cfg**    | 7.0           | 6.0-8.0     |
  | **width**  | 512           | 1024        |
  | **height** | 512           | 1024        |
</details>

## Troubleshooting

### Smart Error Diagnosis System

LobeChat integrates a smart error handling system that can automatically diagnose and provide targeted solutions.

#### Error Types and Solutions

| Error Type         | User Prompt                        | Automatic Diagnosis                                 |
| ------------------ | ---------------------------------- | --------------------------------------------------- |
| **Connection**     | "Cannot connect to ComfyUI server" | Auto-detect server status and connectivity          |
| **Authentication** | "API key invalid or expired"       | Auto-verify authentication credentials              |
| **Permissions**    | "Access permissions insufficient"  | Auto-check user permissions and file access         |
| **Model Issues**   | "Cannot find specified model file" | Auto-scan available models and suggest alternatives |
| **Configuration**  | "Configuration file error"         | Auto-verify config completeness and syntax          |

<details>
  <summary><b>🔍 Traditional Troubleshooting Methods</b></summary>

  #### 1. Connection Failure

  **Issue**: Cannot connect to ComfyUI server

  **Solution**:

  ```bash
  # Confirm server running
  curl http://localhost:8188/system_stats

  # Check port
  netstat -tulpn | grep 8188
  ```

  #### 2. Out of Memory

  **Issue**: Memory errors during generation

  **Solution**:

  - Lower image resolution
  - Reduce generation steps
  - Use quantized models

  #### 3. Authentication Failure

  **Issue**: 401 or 403 errors

  **Solution**:

  - Verify authentication configuration
  - Check if Token is expired
  - Confirm user permissions
</details>

## Best Practices

### Prompt Writing

1. **Detailed Description**: Provide clear, detailed image descriptions
2. **Style Specification**: Clearly specify artistic style, color style, etc.
3. **Quality Keywords**: Add "4K", "high quality", "detailed" keywords
4. **Avoid Contradictions**: Ensure description content is logically consistent

**Example**:

```plaintext
A young woman with flowing long hair, wearing an elegant blue dress, standing in a cherry blossom park,
sunlight filtering through leaves, warm atmosphere, cinematic lighting, 4K high resolution, detailed, photorealistic
```

### Parameter Optimization

1. **FLUX Schnell**: Suitable for quick previews, use 4-step generation
2. **FLUX Dev**: Balance quality and speed, CFG 3.5, 20 steps
3. **FLUX Krea-dev**: Safe creation, CFG 4.5, note content filtering
4. **FLUX Kontext-dev**: Image editing, strength 0.6-0.9

<Callout type={'warning'}>
  Please note during use:

  - FLUX Dev, Krea-dev, Kontext-dev models are for non-commercial use only
  - Generated content must comply with relevant laws and platform policies
  - Large model generation may take considerable time, please be patient
</Callout>

## API Reference

<details>
  <summary><b>📚 API Documentation</b></summary>

  ### Request Format

  ```typescript
  interface ComfyUIRequest {
    model: string; // Model ID, e.g., 'flux-schnell'
    prompt: string; // Text prompt
    width: number; // Image width
    height: number; // Image height
    steps: number; // Generation steps
    seed: number; // Random seed
    cfg?: number; // CFG Scale (Dev/Krea/Kontext specific)
    strength?: number; // Edit strength (Kontext specific)
    imageUrl?: string; // Input image (Kontext specific)
  }
  ```

  ### Response Format

  ```typescript
  interface ComfyUIResponse {
    images: Array<{
      url: string; // Generated image URL
      filename: string; // Filename
      subfolder: string; // Subdirectory
      type: string; // File type
    }>;
    prompt_id: string; // Prompt ID
  }
  ```

  ### Error Codes

  | Error Code | Description              | Resolution Suggestions           |
  | ---------- | ------------------------ | -------------------------------- |
  | `400`      | Invalid parameters       | Check parameter format and range |
  | `401`      | Authentication failed    | Verify API key and auth config   |
  | `403`      | Insufficient permissions | Check user permissions           |
  | `404`      | Model not found          | Confirm model file exists        |
  | `500`      | Server error             | Check ComfyUI logs               |
</details>

You can now use ComfyUI in LobeChat for high-quality AI image generation and editing. If you encounter issues, please refer to the troubleshooting section or consult the [ComfyUI official documentation](https://github.com/comfyanonymous/ComfyUI).