ministudio 0.1.0

Model-Agnostic AI Video Generation Framework

Ministudio

Model-Agnostic AI Video Framework

"Make AI video generation as consistent as CSS makes web styling"

Python package for generating consistent AI videos across different providers.

Installation

pip install ministudio

With provider support:

pip install ministudio[vertex-ai]  # Google Vertex AI
pip install ministudio[openai]     # OpenAI Sora
pip install ministudio[all]        # All providers

Quick Start

Basic Usage

# Mock provider (no API keys needed)
ministudio --provider mock --concept "Hello World" --action "orb waving"

Real Providers

# Google Vertex AI
export GCP_PROJECT_ID="your-project-id"
ministudio --provider vertex-ai --concept "Nature" --action "forest growing"

# OpenAI Sora
export OPENAI_API_KEY="your-key"
ministudio --provider openai-sora --concept "Ocean" --action "waves crashing"

Python API

from ministudio import Ministudio

provider = Ministudio.create_provider("mock")
studio = Ministudio(provider=provider)

result = await studio.generate_concept_video(
    concept="Math",
    action="orb solving equations"
)

Features

• Model-Agnostic: Swap AI providers without changing code
• State Management: Maintain visual consistency across generations
• Configurable: Customize all video generation parameters
• Segmentation: Generate long-form videos with state persistence
• Multiple Providers: Google Vertex AI, OpenAI Sora, Local models, Mock
• Templates: Pre-built styles for different use cases
• API Server: Self-hosted REST API
• Docker: Containerized deployment

Usage

Basic Generation

from ministudio import Ministudio, VideoConfig

provider = Ministudio.create_provider("mock")
studio = Ministudio(provider=provider)

result = await studio.generate_concept_video(
    concept="Science",
    action="orb demonstrating physics"
)

Configurable Generation

from ministudio import VideoConfig

config = VideoConfig(
    duration_seconds=12,
    style_name="cinematic",
    mood="dramatic"
)

result = await studio.generate_concept_video(
    concept="Adventure",
    action="hero exploring cave",
    config=config
)

Segmented Videos (Long-form)

segments = [
    {"concept": "Intro", "action": "character enters scene"},
    {"concept": "Action", "action": "character finds treasure", "state_updates": {"character": {"holding": "treasure"}}},
    {"concept": "Climax", "action": "character escapes"}
]

results = await studio.generate_segmented_video(segments, config)

Providers

Provider	Setup	Status
Mock	None	Ready
Google Vertex AI	GCP_PROJECT_ID env var	Ready
OpenAI Sora	OPENAI_API_KEY env var	Ready
Local	Model path	Ready

Styles & Templates

Built-in Styles

• ghibli: Studio Ghibli aesthetic
• cyberpunk: Neon cyberpunk style
• cinematic: Hollywood filmmaking
• realistic: Photorealistic

Templates

• explainer: Educational content
• marketing: Promotional videos
• cinematic: Cinematic scenes

API Server

Run self-hosted API:

uvicorn ministudio.api:app --host 0.0.0.0 --port 8000

Docker

docker build -t ministudio .
docker run -p 8000:8000 ministudio

API Reference

Ministudio Class

• generate_concept_video(concept, action, config=None)
• generate_segmented_video(segments, base_config=None)

VideoConfig Class

• Configurable parameters: duration, aspect_ratio, style, provider, etc.

Providers

• Ministudio.create_provider(type, **kwargs)

Publishing to PyPI

To publish your package to PyPI:

1. Install build tools:

   pip install build twine
   

2. Build the package:

   python -m build
   

3. Create PyPI account:

   • Go to https://pypi.org/
   • Create account and verify email

4. Upload to PyPI:

   twine upload dist/*
   

   Enter your PyPI username and password when prompted.

5. Test installation:

   pip install your-package-name
   

For test releases, use TestPyPI:

twine upload --repository testpypi dist/*
pip install --index-url https://test.pypi.org/simple/ your-package-name

Philosophy

Ministudio exists because AI video generation is powerful but inconsistent. We believe every developer should be able to create professional videos, AI should enhance creativity when making videos, open ecosystems beat walled gardens, consistency is programmable with proper state management, and the best tools get out of the way.

We're building the standard framework for AI video generation - model-agnostic, stateful, and extensible.

Contributing

We welcome contributions! This guide explains how to get started.

Development Setup

git clone https://github.com/aynaash/ministudio.git
cd ministudio
pip install -e .[all]

Code Quality

# Format
black ministudio/

# Lint  
ruff check ministudio/

# Test
pytest

Adding Providers

Extend BaseVideoProvider and implement generate_video() method.

Pull Requests

• Reference issues
• Clear descriptions
• Include tests
• Follow code style

Thank you for contributing to Ministudio!

License

MIT License

Acknowledgments

Inspired by the open-source AI community's work on making AI accessible and consistent.

---

Made with by the AI video generation community

Ready to make AI video generation consistent? Let's build the future together.