SDK Documentation
Reality Defender SDK
The Reality Defender SDK provides tools and libraries for detecting deepfakes and manipulated media through the Reality Defender API. This SDK is available in multiple programming languages to fit your development needs.
Overview
The Reality Defender SDK allows you to integrate powerful AI-based deepfake detection capabilities into your applications. With this SDK, you can:- Upload media files for deepfake and manipulation analysis
- Receive detailed results about the authenticity of media
- Get model-specific confidence scores and detection results
- Integrate via event-based or polling approaches
- Process multiple files concurrently with configurable concurrency limits
- Handle both image, video, audio and text files with optimized processing
Available SDKs
SDK implementations are available for multiple programming languages:- TypeScript/JavaScript SDK - For web and Node.js applications
- Python SDK - For Python applications and data science workflows
- Go SDK - For Go applications
- Rust SDK - For Rust applications
- Java SDK - For Java applications
Getting Started
- Obtain an API key from the Reality Defender Platform
- Choose the SDK for your preferred programming language
- Follow the installation and usage instructions in the language-specific README
Supported File Types
- Documents:
.pdf,.doc,.docx,.txt - Images:
.jpg,.jpeg,.png,.gif,.webp - Audio:
.mp3,.wav,.m4a,.aac,.ogg,.flac,.alac - Video:
.mp4,.mov - Text:
.txt
Size Limits
- Documents: up to 5MB
- Images: up to 10MB
- Audio: up to 20MB
- Video: up to 250MB
Architecture
The SDKs follow a consistent architecture across all language implementations:- Client Layer: Handles HTTP communication with the Reality Defender API
- Core: Manages configuration, constants, and event handling
- Detection: Processes media uploads and results
- Types/Models: Defines data structures for API responses and SDK interfaces
- Utils: Provides file operations and helper functions
Key Features
- Cross-language compatibility: Consistent API across TypeScript, Python, Go, and Rust
- Async/Sync support: Both asynchronous and synchronous programming models
- Score normalization: All scores are normalized to a 0-1 range (0.0 to 1.0)
- Resource management: Proper cleanup of resources to prevent leaks
- Flexible integration: Event-based or polling-based approaches
- Batch processing: Process multiple files concurrently with optimized performance
- Media type support: Handle audio, image, video and text files with appropriate processing strategies