Project Montage is a modular, AI-powered video builder that enables developers to rapidly create fully edited, production-quality videos directly from code. Built on Google's advanced generative AI ecosystem—powered by Gemini's orchestration capabilities and integrated with Veo and Nano Banana models—it transforms a simple storyboard outline into a cohesive narrative video, complete with cinematic clips, transitions, and generative music.

Video production with multiple AI models often involves complex orchestration and fragmented workflows. Project Montage addresses this by providing a unified, agentic pipeline that automates the entire creative lifecycle. Here are the key benefits:

• Unrivaled Flexibility & Seamless Enterprise Data Integration: Streamline internal data into high-volume marketing video production.
• Multi-Agent Synergy & High-Fidelity Content Output: Create cinema-quality, natively synced videos using Google's premier AI models (Nano Banana, Veo).
• Open-Source Infrastructure for Rapid, Ready-to-Deploy Implementation: Deploy a ready-to-use framework instantly while retaining technical independence.

Project Montage is designed with a highly modular multi-agent framework. While you can modify the code to integrate other third-party services, the project is optimized for Google Cloud Platform (GCP) out-of-the-box, leveraging powerful models like Gemini and Veo for the smooth experience. We highly recommend setting up GCP to get your first video running smoothly.

• Go to the Google Cloud Console.
• Create a new project or select an existing one.
• Important: Ensure that Billing is enabled for your Cloud project to use Vertex AI.

Navigate to Vertex AI in the console and click "Enable All Recommended APIs," or run the following in your terminal:

gcloud services enable aiplatform.googleapis.com

For local development, use Application Default Credentials (ADC). Install the Google Cloud CLI if you haven't already, then run:

# Login with your Google account
gcloud auth application-default login

# Set your active project
gcloud config set project <YOUR_PROJECT_ID>

Get your first video generated in minutes. After cloning the repository, set up and run the core services in three separate terminals:

# First, clone the project
git clone https://github.com/google/project-montage.git
cd project-montage

# --- Terminal 1: Start the MCP Server ---
cd mcp_montage
uv sync
bash sync_fonts_from_third_party.sh
cp .env.example .env # Edit .env with your configuration
uv run uvicorn server:app --port 8001 --reload

# --- Terminal 2: Start the MCP Client (Web UI) ---
cd mcp_client
uv sync
cp .env.example .env # Edit .env with your configuration
uv run uvicorn server:app --port 8000 --reload

# --- Terminal 3: Start the Sign Server ---
cd sign_server
uv sync
cp .env.example .env # Edit .env with your configuration
uv run uvicorn server:app --port 8080

Once running, open the ADK Web UI in your browser to orchestrate the video production workflow!

• AI-Powered Orchestration: A central agent powered by Gemini interprets your creative outline and automatically assigns tasks to specialized agents.
• Multi-Agent Architecture:
  • Storyboard Generation: Automatically drafts a scene-by-scene plan and selects optimal input images.
  • Intelligent Asset Selection: Automatically decides which input images best match generated scene descriptions.
  • Cinematic Video Generation: Leverages Google's Veo 3.1 model to create high-quality, dynamic video clips.
  • Character Integration: Uses Nano Banana for consistent character integration and asset resizing.
• Automated Post-Production: Seamlessly stitches clips together with professional transitions and generative background music aligned to the content theme.

(See Montage in action! Orchestrates multi-agent workflows to transform a storyboard outline into a fully-edited, professional video.)

We welcome contributions! Please set up the environment using the following steps:

# Set up pre-commit hooks
pip install pre-commit && pre-commit install

# Lint and Format with Ruff
uvx ruff check .
uvx ruff format .

# Run Tests
pytest tests/

Please open an issue first to discuss any significant changes before submitting a Pull Request.