A production-ready, containerized web server that connects to Azure OpenAI's Sora 2 model to generate videos. This application provides both a modern web interface and RESTful API endpoints for video generation.

• Web Interface: Intuitive HTML/CSS/JavaScript frontend for easy video generation
• Enhanced Video History: Advanced history page with search, filtering, sorting, and modal video playback
  • 🔍 Search: Filter videos by prompt text
  • 📊 Sort: Sort by date, resolution, or duration
  • 🎯 Filter: Filter by resolution, duration, and status
  • 🎬 Modal Playback: Click any video to view in a full-screen modal
  • 🗑️ Delete: Remove videos and their history entries
• Persistent Storage: Videos are downloaded and saved locally with volume mount support
• Azure OpenAI Integration: Seamless connection to Azure OpenAI Sora 2 video generation
• Containerized: Multi-stage Docker build optimized for ease of use

• Python 3.11+
• Docker (for containerized deployment)
• Azure OpenAI account with Sora 2 access
• Azure OpenAI API key and endpoint

The service supports two configuration modes:

Use AZURE_OPENAI_VIDEO_URL to specify the complete endpoint URL. This is the most flexible approach and works with different Azure services and URL structures:

export AZURE_OPENAI_API_KEY="your-api-key"
export AZURE_OPENAI_VIDEO_URL="https://your-instance.cognitiveservices.azure.com/openai/v1/videos"
export AZURE_OPENAI_DEPLOYMENT="sora-2"

This mode is recommended because:

• It gives you full control over the endpoint URL
• It works with both .openai.azure.com and .cognitiveservices.azure.com domains
• It supports different API path structures
• It allows you to omit API versions if not required by your service

Use AZURE_OPENAI_ENDPOINT to let the service construct the URL automatically:

export AZURE_OPENAI_API_KEY="your-api-key"
export AZURE_OPENAI_ENDPOINT="https://your-instance.openai.azure.com/"
export AZURE_OPENAI_DEPLOYMENT="sora-2"
export AZURE_OPENAI_API_VERSION="2024-08-01-preview"

This mode automatically constructs URLs in the format: {ENDPOINT}/openai/deployments/{DEPLOYMENT}/videos?api-version={API_VERSION}

1. Build the Docker image

   docker build -t azure-openai-sora-webserver .

2. Run the container

   Option A: Using Custom Video URL (Recommended)

   docker run -d \
     --name sora-webserver \
     -p 8000:8000 \
     -v $(pwd)/data:/app/data \
     -e AZURE_OPENAI_API_KEY="your-api-key" \
     -e AZURE_OPENAI_VIDEO_URL="https://your-instance.cognitiveservices.azure.com/openai/v1/videos" \
     -e TZ="America/New_York" \
     azure-openai-sora-webserver

   Option B: Using Legacy Endpoint Configuration

   docker run -d \
     --name sora-webserver \
     -p 8000:8000 \
     -v $(pwd)/data:/app/data \
     -e AZURE_OPENAI_API_KEY="your-api-key" \
     -e AZURE_OPENAI_ENDPOINT="https://your-instance.openai.azure.com/" \
     -e TZ="America/New_York" \
     azure-openai-sora-webserver

   Notes:

   • The -v $(pwd)/data:/app/data mount creates a persistent volume for storing generated videos and history
   • Set TZ to your local timezone for accurate log timestamps (e.g., America/New_York, Europe/London, Asia/Tokyo). Defaults to UTC if not specified.
   • Volume Permissions: The container automatically fixes permissions on mounted volumes. If the mounted directory is owned by root or another user, the container will change ownership to the internal appuser at startup. This ensures the application can write videos to the mounted volume without permission errors.

3. Access the application

   • Web Interface: http://localhost:8000
   • Health Check: http://localhost:8000/health

The application is automatically built and published to GitHub Container Registry (GHCR):

Option A: Using Custom Video URL (Recommended)

docker pull ghcr.io/loryanstrant/azure-openai-sora-webserver:latest
docker run -d \
  --name sora-webserver \
  -p 8000:8000 \
  -v $(pwd)/data:/app/data \
  -e AZURE_OPENAI_API_KEY="your-api-key" \
  -e AZURE_OPENAI_VIDEO_URL="https://your-instance.cognitiveservices.azure.com/openai/v1/videos" \
  -e TZ="America/New_York" \
  ghcr.io/loryanstrant/azure-openai-sora-webserver:latest

Option B: Using Legacy Endpoint Configuration

docker pull ghcr.io/loryanstrant/azure-openai-sora-webserver:latest
docker run -d \
  --name sora-webserver \
  -p 8000:8000 \
  -v $(pwd)/data:/app/data \
  -e AZURE_OPENAI_API_KEY="your-api-key" \
  -e AZURE_OPENAI_ENDPOINT="https://your-instance.openai.azure.com/" \
  -e TZ="America/New_York" \
  ghcr.io/loryanstrant/azure-openai-sora-webserver:latest

1. Navigate to http://localhost:8000
2. Enter your video description in the prompt field
3. Select resolution (1280x720 landscape or 720x1280 portrait)
4. Set duration (4, 8, or 12 seconds)
5. (Optional) Upload a reference image for image-to-video generation
6. Click "Generate Video"
7. Monitor progress and view the generated video
8. View all generated videos at http://localhost:8000/static/history.html

The container runs as root by default, which eliminates most permission issues when mounting volumes. The entrypoint script automatically:

1. Creates necessary directories (/app/data and /app/data/videos)
2. Ensures directories are accessible and writable
3. Starts the application with full permissions

If you prefer to run the container with a specific user ID (not recommended), you can use:

docker run -d \
  --user $(id -u):$(id -g) \
  -v /path/to/data:/app/data \
  ...other options...
  azure-openai-sora-webserver

Note: Running as a non-root user may cause permission issues with mounted volumes.

Verification: Check the container logs to verify startup:

docker logs your-container-name

You should see output like:

Directory /app/data exists
Directory /app/data is writable
Starting application as root...

This project is licensed under the MIT License - see the LICENSE file for details.

• Azure OpenAI for providing Sora 2 video generation capabilities
• FastAPI for the excellent web framework
• The Python community for outstanding tooling and libraries

This application uses the Azure OpenAI Sora 2 API, which provides the following features:

• Model: sora-2 (configurable via AZURE_OPENAI_DEPLOYMENT)
• Video Durations: 4, 8, or 12 seconds per generation
• Resolutions:
  • Landscape: 1280x720
  • Portrait: 720x1280
• Generation Modes:
  • Text-to-Video: Generate videos from text prompts
  • Image-to-Video: Animate still images by providing an input reference image
• Input Image Requirements:
  • Supported formats: JPEG, PNG, WebP
  • Resolution must match the selected video resolution exactly
  • Used as a visual anchor for the first frame
• API Endpoints:
  • client.videos.create() - Start video generation
  • client.videos.retrieve() - Check generation status
  • client.videos.download_content() - Download completed video
• Status Values: queued, in_progress, completed, failed, cancelled

For more information, see the Microsoft documentation.