google/magika
Fast and accurate AI powered file content types detection
Employs a lightweight deep-learning model (~few MBs) trained on ~100M files across 200+ content types, achieving ~99% accuracy with near-constant 5ms inference regardless of file size. Features per-content-type confidence thresholds and adjustable prediction modes (high/medium/best-guess) to control error tolerance. Available as a Rust CLI, Python API, and bindings for JavaScript/TypeScript and Go, with production deployment at Google processing hundreds of billions of files weekly.
10,151 stars and 5,346,208 monthly downloads. Used by 3 other packages. Actively maintained with 11 commits in the last 30 days. Available on PyPI.
Stars
10,151
Forks
495
Language
Python
License
Apache-2.0
Category
Last pushed
Mar 03, 2026
Monthly downloads
5,346,208
Commits (30d)
11
Dependencies
2
Reverse dependents
3
Get this data via API
curl "https://pt-edge.onrender.com/api/v1/quality/ml-frameworks/google/magika"
Open to everyone — 100 requests/day, no key needed. Get a free key for 1,000/day.
Related frameworks
meilfang/LMFD-PAD
Learnable Multi-level Frequency Decomposition and Hierarchical Attention Mechanism for...
athen-lab/mai
Multilayer Authenticity Identifier (MAI), a CNN model that attempts to identify synthetic AI images.
manjaryp/GANvsGraphicsvsReal
Distinguishing Natural and Computer-Generated Images using Multi-Colorspace fused EfficientNet
2spi/ai-v-real
Real/AI Generated Image Classifier
Saranya-T-S/AI-Image-Detector
Deep learning-based system to detect AI-generated images using ELA, PRNU, FFT, and noise...