Metis Halo
Cloud-based profanity detection and cleaning for websites, keeping user-generated content appropriate with nothing to install or maintain.
Metis Halo is Helium22's profanity detection and cleaning service for websites, delivered from the cloud so you can keep user-generated content clean without installing or maintaining anything yourself. Wherever your audience contributes text, Metis Halo helps you keep it appropriate.
Online communities, comment sections, reviews, forms and messaging all carry the risk of offensive or inappropriate language. Left unchecked, this damages your brand, harms your users' experience, and can create real compliance and safeguarding concerns.
Metis Halo screens content in real time, identifying profanity and unwanted language and cleaning or flagging it according to your rules. Because it runs in the cloud, it scales effortlessly with your traffic and is always up to date.
The service is designed to be flexible. You decide how strict to be, which terms matter in your context, and whether content should be filtered, masked or held for review, so moderation reflects your community's standards rather than a one-size-fits-all setting.
Integration is straightforward, allowing you to add robust language moderation to your existing website or application without a heavy engineering effort or ongoing infrastructure to manage.
With Metis Halo protecting your platform, you can offer open, interactive experiences to your users while keeping your content safe, respectful and on-brand.
How Metis Halo works
Metis uses AI-powered detection to find and clean profanity — including obfuscated, reversed and fuzzy-matched variants that simple filters miss.
Register & get API keys
Create your free account to receive a unique Tenant Key and API Key. These authenticate every request to the Metis scanning engine.
Submit content for scanning
Send text, upload files (TXT, CSV, JSON, XML) or connect a database. Each scan uses one credit from your balance.
AI-powered detection
The engine analyses content using fuzzy matching, reverse detection, obfuscation decoding and multi-language support to catch what simpler filters miss.
Get detailed reports
Comprehensive reports show every match with exact positions, match types, severity levels and confidence scores. Fetch reports by ID for audit trails.
Optional cleaning & redaction
Turn on cleaning to redact detected profanity automatically. Download cleaned files or receive sanitised text with configurable replacement characters.
Detection capabilities
| Match type | Description | Example |
|---|---|---|
| Exact match | Direct dictionary lookup for known profane terms across multiple languages. |
badword
|
| Fuzzy match | Detects misspellings and close variants using Levenshtein distance and phonetic matching. |
bádw0rd, badwrod
|
| Obfuscation | Identifies character substitution: leetspeak, Unicode homoglyphs and symbol insertion. |
b@dw0rd, b.a.d.w.o.r.d
|
| Reverse detection | Catches words written backwards to evade simple filters. |
drowdab
|
| Multi-language | Detects profanity across multiple languages and character sets simultaneously. |
Various languages
|
| Context-aware | Reduces false positives by considering word context and compound word boundaries. |
Scunthorpe → safe
|
Built for integration. Every scan is available through a simple authenticated API, so Metis can sit quietly behind your website, app or data pipeline and keep user-generated content clean without anyone having to police it by hand.