Sora Custom Node for ComfyUI

This repository contains a comprehensive set of ComfyUI custom nodes for interacting with a Sora-compatible REST API. The nodes support the full range of Sora capabilities:

• Text-to-Video: Generate videos from text prompts
• Image-to-Video: Generate videos from images
• Video-to-Video: Edit, extend, or transform existing videos

All nodes use only Python's standard library (with optional PIL/numpy support for image processing).

Installation

1. Clone or copy this folder into the ComfyUI/custom_nodes/ directory:

   cd ComfyUI/custom_nodes
   git clone <this repo> sora
   

   重要: 确保文件夹结构如下：

   ComfyUI/custom_nodes/sora/
   ├── __init__.py          ← 必需！
   ├── sora_node.py         ← 必需！
   └── ...
   

2. 完全重启 ComfyUI（关闭所有进程，然后重新启动）

   • 不是刷新浏览器页面
   • 需要完全关闭并重新启动 ComfyUI 服务器

3. Optional dependencies (for better image handling):

   • Pillow (PIL): For processing PIL Image objects
   • numpy: For processing numpy array inputs

   These are optional - the nodes will work with file paths and base64 strings without them.

Verification

After installation, verify the nodes are loaded:

cd ComfyUI/custom_nodes/sora
python test_comfyui_load.py

If nodes don't appear in ComfyUI, see TROUBLESHOOTING.md for help.

Available Nodes

1. Sora Text-To-Video

Generates videos from text prompts.

Inputs:

• prompt (required): Text description of the video
• negative_prompt: What to avoid in the video
• duration_seconds: Video length (default: 5.0)
• aspect_ratio: Video aspect ratio (default: "16:9")
• fps: Frames per second (default: 24)
• guidance_scale: Generation guidance (default: 7.5)
• seed: Random seed for reproducibility
• And more...

2. Sora Image-To-Video

Generates videos from images. Accepts ComfyUI image inputs (PIL Image, numpy arrays, or file paths).

Inputs:

• image (required): Image input (from ComfyUI image nodes)
• prompt (required): Text description for video generation
• All the same optional parameters as Text-To-Video

Image Input Support:

• PIL Image objects
• Numpy arrays (various shapes supported)
• File paths (PNG, JPG, WebP)
• Base64 encoded strings

3. Sora Video-To-Video

Edits or extends existing videos. Supports multiple operations:

• extend: Extend video duration
• edit: Edit video content based on prompt
• inpaint: Fill masked regions
• style_transfer: Apply style to video

Inputs:

• video (required): Video file path or base64 string
• prompt (required): Description of desired changes
• operation: Type of video operation (default: "extend")
• extension_seconds: For extend operation (default: 5.0)
• All the same optional parameters as other nodes

Configuration

All nodes look for an API key using the following priority:

1. The api_key input field on the node.
2. The SORA_API_KEY environment variable.
3. The OPENAI_API_KEY environment variable.

You can also configure:

• base_url: defaults to https://api.openai.com/v1/sora.
• endpoint: defaults to /videos, change it if your backend uses another path.
• download_path: optional local directory or file path for saving the video.
• wait_for_result: Whether to poll until completion (default: True)
• poll_interval: Seconds between status checks (default: 3.0)
• max_wait_seconds: Maximum wait time (default: 120.0)

Outputs

All nodes return three strings:

1. Job ID: Unique identifier for the generation job
2. Status: Current status (submitted, processing, succeeded, failed, etc.)
3. Result JSON: Contains the raw API response and downloaded file path (if any)

Local Testing

You can test the custom node locally without running ComfyUI:

1. Basic Functionality Test

Run the test script to verify node functionality:

# Run basic tests (no API calls)
python test_sora_node.py

# Run full unittest suite
python test_sora_node.py --unittest

# Test with real API (requires API key)
python test_sora_node.py --test-client --api-key YOUR_KEY

2. Test in ComfyUI

1. Copy this folder to ComfyUI/custom_nodes/sora/
2. Start ComfyUI
3. The node should appear in the node menu under "Sora" category
4. You can test it directly in the ComfyUI interface

3. Mock Testing

The test script uses mocks to test API interactions without making real network calls. This is useful for:

• Testing error handling
• Testing response parsing
• Validating node logic

Features

Full Sora Capability Support

• ✅ Text-to-Video generation
• ✅ Image-to-Video generation
• ✅ Video-to-Video editing and extension
• ✅ Multiple video operations (extend, edit, inpaint, style transfer)

ComfyUI Integration

• Seamless integration with ComfyUI image nodes
• Supports all ComfyUI image formats (PIL, numpy, file paths)
• Automatic format conversion and handling

Robust Error Handling

• Clear error messages for API failures
• Timeout protection
• Automatic retry logic for polling

Flexible Configuration

• Environment variable support for API keys
• Customizable endpoints and base URLs
• Optional result downloading
• Webhook support for async operations

Notes

• Nodes poll the job until completion when wait_for_result is enabled.
• If download_path is a folder, videos are saved as <job_id>.mp4.
• The nodes automatically search for video URLs in various response formats:
  • result_url or video_url (direct)
  • outputs[0].url or outputs[0].download_url (nested)
  • videos array
• Image inputs are automatically converted to the format required by the API.
• Video inputs support file paths and base64 encoded strings.