# Tabu

> AI-powered explicit content moderation API that classifies images and videos across 5 categories with sub-200ms latency to help apps pass App Store UGC reviews and meet DSA compliance.

Tabu is an AI-powered content moderation API built by Producon s.r.o. that detects and blocks explicit images and videos in milliseconds. It targets developers and founders who need to pass Apple App Store Guideline 1.2 UGC reviews, automate EU Digital Services Act compliance, and protect users from nudity and pornographic content. The API is powered by the open-source MobileNetV2 architecture and processes media entirely in memory with zero data retention by default.

## What It Is

Tabu is a REST API service that classifies images and videos into five categories — Neutral, Sexy, Porn, Hentai, and Drawing — returning confidence scores for each. Developers send a single POST request with an image URL or file upload and receive a structured JSON response in under 200ms on average. The service sits in front of user-generated content upload flows, acting as a real-time gate that blocks explicit media before it reaches storage or other users.

## How the Classification Pipeline Works

The core classification endpoint (`/v1/classify`) accepts images by URL or multipart file upload. For video, a separate `/v1/classify/video` endpoint extracts one frame every three seconds and processes up to approximately three minutes of footage synchronously. Results include a `safe` boolean, a `main_category` label, a top confidence score, and a full breakdown of all five category scores. In-memory SHA-256 caching means repeated submissions of the same file return in under 10ms without re-running inference.

## Compliance and Human-in-the-Loop Features

Beyond raw classification, Tabu includes a dashboard-based review queue where human moderators can override AI decisions with a single click. Webhook notifications push override events to the developer's own backend in real time, enabling platforms to restore or permanently remove content based on human judgment. The service also auto-generates downloadable PDF transparency reports designed to satisfy EU DSA auditable moderation requirements and demonstrate "reasonable care" to regulators. Configurable sensitivity sliders for Porn, Hentai, and Sexy categories — independently for images and video — let teams tune false-positive rates without code changes.

## Developer Setup Path

Tabu is designed for a sub-five-minute integration. The homepage provides code samples in cURL, Node.js, and Python. An open-source Express.js boilerplate on GitHub (`sulejj/tabu-nsfw`) provides a drop-in backend route using Express and Multer that intercepts uploads and calls the API automatically. The boilerplate targets social networks, dating apps, community forums, and chat bots on platforms like Discord, Slack, and Telegram. No credit card is required to start, and API keys are generated from the dashboard.

## Privacy Architecture

By default, Tabu operates in Privacy-First Mode: images and videos are processed entirely in server memory and instantly destroyed after classification. Only text metadata — scores, timestamps, and request IDs — is retained in logs. Developers can optionally disable this mode to build a visual audit log in the dashboard or support future AI fine-tuning, but the docs note this requires updating the application's privacy policy to disclose third-party content processing.

## Current Status

The GitHub boilerplate repository (`sulejj/tabu-nsfw`) was created on 2026-08-18 and last updated on 2026-08-19, indicating a very recent launch. The service is live and accepting signups with no credit card required on the free tier.

## Features
- Image classification via URL or file upload
- Video classification (1 frame/3 seconds, up to ~3 min)
- 5-category classification: Neutral, Sexy, Porn, Hentai, Drawing
- Sub-200ms average latency
- 93%+ accuracy
- Configurable sensitivity thresholds per category
- Human-in-the-loop review queue
- Webhook notifications for admin overrides
- One-click DSA transparency PDF reports
- Privacy-First Mode (zero data retention)
- In-memory SHA-256 caching for repeated submissions
- Dashboard with API key management
- Express.js boilerplate for drop-in integration

## Integrations
cURL, Node.js, Python, Express.js, Multer, Discord, Slack, Telegram, React Native, Next.js

## Platforms
IOS, WEB, API

## Pricing
Freemium — Free tier available with paid upgrades

## Links
- Website: https://tabushield.com
- Documentation: https://tabushield.com/docs/
- Repository: https://github.com/sulejj/tabu-nsfw
- EveryDev.ai: https://www.everydev.ai/tools/tabu