Architecture
Understanding the core architecture of MiniStudio
Overview
MiniStudio is built on a modular, provider-agnostic architecture that allows you to compose complex video generation workflows with state management, character consistency, and temporal continuity.
VideoOrchestrator
The main orchestration engine that manages video generation workflows, state transitions, and result aggregation.
Provider Interface
Abstraction layer supporting multiple video generation backends (Vertex AI, Sora, custom providers).
State Machine
Manages continuity between shots using the Invisible Weave algorithm for environment and character consistency.
Identity Grounding
Ensures character consistency across multiple shots using Identity Grounding 2.0 technology.
Key Components
VideoGenerationRequest
Encapsulates a single video generation request with prompt, duration, aspect ratio, and optional state parameters.
VideoGenerationResult
Contains the generated video path, metadata, frame data for continuity, and any state information for multi-shot sequences.
Provider
Abstract base class that all video generation providers must implement. Standardizes how requests are sent and results are returned.
Data Flow
Design Principles
- •Modularity: Each component has a single responsibility and can be extended independently.
- •Provider Agnosticism: Works with any video generation API through the Provider interface.
- •State Management: Built-in state machine for maintaining continuity across shots.
- •Async-First: All operations are async-compatible for high-performance workflows.