openmoderate¶

Open-source, self-hosted image and video content moderation for Python. Detect nudity, sexual content, violence, gore, weapons, hate symbols, drugs, self-harm, alcohol and tobacco in uploaded images and videos — with one pip install, no cloud account, no per-image fees, and no ML experience needed.

Think of it as an open alternative to Amazon Rekognition's DetectModerationLabels that runs inside your own app.

Terminal demo: openmoderate scans a folder, flagging a pistol, a riot and a swastika while a beach, a dog and a US flag pass as safe; then the REST server flags a cigarette photo as tobacco.

Install¶

pip install openmoderate

Requires Python 3.9+. The first call downloads the CLIP model weights (~600 MB) and caches them; after that everything runs offline. CPU works fine; CUDA and Apple Silicon (MPS) are used automatically when available.

On a CPU-only Linux server you can save ~2 GB by installing the CPU build of PyTorch first. Install torch and torchvision together from the same index, otherwise their versions won't match:

pip install torch torchvision --index-url https://download.pytorch.org/whl/cpu
pip install openmoderate

Optional extras: openmoderate[server] (REST API), [video], [fastapi], [flask], [django].

Quick start¶

from openmoderate import Moderator

mod = Moderator()
result = mod.analyze("upload.jpg")

if not result.is_safe:
    print(result.violations)
    # [Label(name='weapons', confidence=0.9612, parent='violence', threshold=0.5)]

How openmoderate works

Where to go next¶

I want to…	Read
Understand how scoring works and what to expect	How it works
Call it from Python and tune the policy	Python API
Scan folders from a terminal or CI	Command line
Improve accuracy with my own images	Train on your own data
Moderate videos	Videos
Protect uploads in FastAPI, Flask or Django	Web framework integrations
Call it over HTTP from another language	REST API server
See measured accuracy	Benchmarks

Versioning¶

openmoderate follows semantic versioning from 1.0: everything importable from the top-level openmoderate package, the integration modules, the CLI flags, the REST endpoints and the .omh head format are stable within a major version. Default prompts, thresholds and the scores they produce may be tuned in minor releases to improve accuracy; such changes are always listed in the changelog. Pin the minor version if you need bit-identical scores.