Profanity Filter API: 26 Languages, Leetspeak, Allowlist avatar

Profanity Filter API: 26 Languages, Leetspeak, Allowlist

Pricing

from $0.20 / 1,000 text checkeds

Go to Apify Store
Profanity Filter API: 26 Languages, Leetspeak, Allowlist

Profanity Filter API: 26 Languages, Leetspeak, Allowlist

Find and mask bad words in comments, reviews and chat in 26 languages. Catches sh1t, f u c k, fuuuck and f*ck without flagging Password or Scunthorpe. Mask, replace, remove or flag. Bulk runs or an instant API, PurgoMalum compatible.

Pricing

from $0.20 / 1,000 text checkeds

Rating

0.0

(0)

Developer

Pandapi

Pandapi

Maintained by Community

Actor stats

0

Bookmarked

1

Total users

0

Monthly active users

35 minutes ago

Last modified

Features

Limited permissionsSingle-tenantServer interface

Share

Profanity Filter API: 26 Languages, Leetspeak & Allowlist

Find and clean bad words in comments, reviews, chat messages, usernames and any user text, in 26 languages. It catches disguised swearing (sh1t, $hit, f u c k, fuuuck, f*ck, look-alike letters from other alphabets) without censoring innocent words like Password, class, Scunthorpe, cocktail or butter. Mask, replace, remove or just flag. Run it in bulk on a list or a whole dataset, or call it as an instant HTTP API.

  • 🌍 26 languages: English, Spanish, French, German, Italian, Portuguese, Dutch, Polish, Russian, Turkish, Arabic, Persian, Hindi, Chinese, Japanese, Korean, Thai, Swedish, Norwegian, Danish, Finnish, Czech, Hungarian, Filipino, Esperanto, Kabyle
  • 🕵️ Catches disguised words: leetspeak, spaced letters, stretched letters, self-censored words (f**king) and Cyrillic/Greek look-alike letters
  • ✅ No Scunthorpe problem: whole-word matching, tested against 370,105 English dictionary words, so assassin, Essex, spicy or titter are never flagged
  • ✂️ Five modes: mask (****), mask but keep the first letter (f***), replace ([censored]), remove the word, or flag only
  • 🎚️ Two strictness levels: standard (swearing, slurs, explicit sexual slang) or strict (adds mild words like damn, crap, hell and clinical terms, for children's apps)
  • ➕ Your own block and allow lists: frak* blocks every word starting with frak; allow your brand, a place or a surname
  • 🔁 Drop-in PurgoMalum replacement: the same /service/json, /service/plain, /service/xml and /service/containsprofanity endpoints and parameters, so moving over is a base-URL change
  • 🧾 Clean another dataset: point it at the reviews from an App Store or Google Play scraper run and get every item back with a cleaned text field
  • 💸 $0.20 per 1,000 texts. Empty texts are free. No start fee.

Use cases

  • Comment and chat moderation: mask swearing before it is shown, or flag it for a human
  • Usernames and display names: block offensive sign-up names, including sh1tlord-style spellings
  • Reviews and app store data: clean scraped reviews before you put them in a report, a dashboard or an LLM prompt
  • Kids' and education apps: strict mode also catches mild words
  • Datasets for AI training: flag or remove profanity in text corpora in bulk
  • Multilingual communities: one filter for English, Spanish, Russian, Arabic, Japanese and more

How to use it

  1. Paste your texts into Texts to check (one per line), or give the dataset ID of another run and the field that holds the text.
  2. Choose Languages. auto = English plus the language of any non-Latin script found in a text (Russian, Arabic, Persian, Chinese, Japanese, Korean, Hindi, Thai). Add Latin-script languages yourself, e.g. ["en", "es", "fr", "de"], or all.
  3. Pick a mode and run. One output row per text, in the order you gave them. Export to Excel, CSV or JSON.

Example input

{
"texts": [
"What a great product, thanks!",
"This is sh1t and the support is f u c k i n g useless",
"Your Password was reset. Greetings from Scunthorpe."
],
"languages": ["auto"],
"mode": "mask"
}

Example output

{
"id": "2",
"text": "This is sh1t and the support is f u c k i n g useless",
"hasProfanity": true,
"profanityCount": 2,
"cleanText": "This is **** and the support is * * * * * * * useless",
"matches": [
{ "term": "shit", "matched": "sh1t", "start": 8, "end": 12, "language": "en" },
{ "term": "fucking", "matched": "f u c k i n g", "start": 32, "end": 45, "language": "en" }
],
"status": "ok"
}

matches tells you what was found, how it was written, where it is (character offsets) and which language list matched. A run summary (texts checked, texts with profanity, languages) is saved in the key-value store as SUMMARY.

Clean a dataset from another Actor

{ "datasetId": "<dataset ID of an App Store reviews run>", "textField": "text", "mode": "mask-keep-first" }

Every item comes back with all its fields plus textClean, hasProfanity, profanityCount and matches. Nested fields work with dots, e.g. review.body.

Instant API (Standby)

Call it like any REST API, one text or up to 1,000 per request. Use your Apify API token.

# one text
curl "https://blueberry-rainstorm--profanity-filter.apify.actor/check?text=you%20sh1t%20head&token=$APIFY_TOKEN"
# many texts, other options
curl -X POST "https://blueberry-rainstorm--profanity-filter.apify.actor/check?token=$APIFY_TOKEN" \
-H "Content-Type: application/json" \
-d '{"texts": ["ok", "f*ck this"], "mode": "mask-keep-first", "languages": "en,es"}'
# PurgoMalum-compatible
curl "https://blueberry-rainstorm--profanity-filter.apify.actor/service/json?text=this%20is%20shit&fill_text=%5Bx%5D&token=$APIFY_TOKEN"
# {"result":"this is [x]"}

Parameters for /check: text or texts, languages (comma list, auto, all), mode, maskChar, replacement, strictness (standard/strict), obfuscation (true/false), block and allow (comma lists). PurgoMalum endpoints take text, add, fill_text, fill_char, plus our languages and allow. /languages lists the language codes.

Pricing

$0.20 per 1,000 texts checked (event text, $0.0002 each). Empty texts are free, and there is no start fee. 100,000 comments cost $20. Set a maximum cost per run in Apify and the run stops cleanly when it is reached.

How it compares

This ActorPurgoMalumNeutrino Bad Word FilterOther Apify filters
Languages26Englishmostly English1 to 4
Leetspeak, spaced, stretched, look-alike letters✅some substitutions❌some
Allowlist + whole-word matching✅built-in safe list only❌✅
Strictness levelsstandard / strict❌strict / obscenecategories
Bulk dataset cleaning✅❌❌✅
Real-time HTTP API✅ (also PurgoMalum-compatible)✅✅❌
Price$0.20 / 1,000free, no SLApaid plans$0.25 to $2 / 1,000

FAQ

Will it flag words like "Password", "Scunthorpe" or "assassin"? No. It matches whole words after normalizing them, never raw substrings. We ran it over 370,105 English dictionary words and fixed every innocent word that came up (for example butter, spicy, cocktail, Middlesex).

Is this AI moderation? No. It is a fast, predictable word filter: the same input always gives the same output, with the exact word and position. It does not judge tone, sarcasm or insults made of ordinary words. Use it as the first, cheap step, and send flagged texts to a human or an LLM if you need judgement.

Why isn't "damn" or "hell" flagged? They are mild, and most communities allow them. Set Strictness to strict to catch them, along with clinical or everyday words that the full lists include (sexual, sucks, nude).

Which languages does auto check? English always, plus Russian for Cyrillic text, Arabic and Persian for Arabic script, Chinese and Japanese for Han characters, Japanese for kana, Korean for Hangul, Hindi for Devanagari, and Thai for Thai script. Latin-script languages (Spanish, French, German, ...) are not guessed, because a word that is rude in one is often normal in another (Swedish fart means speed). Add them to Languages when your users write in them.

How good are the non-English lists? English and Russian have our own additions and the English list was tuned against a dictionary. The other languages use the community LDNOOBW lists as they are. For Chinese, Japanese and Thai (no spaces between words), terms are matched inside the text, and one-character terms are only used in strict mode to avoid false hits like 牛乳 (milk). Add words to Extra words to block or Words to allow as you need, and tell us in the Issues tab so we can improve the lists.

Do you store my texts? No. Texts are processed in memory. Batch results go to your own Apify dataset, and the API returns them in the response.

Credits

Word lists: List of Dirty, Naughty, Obscene, and Otherwise Bad Words (LDNOOBW), CC BY 4.0, with additions and a dictionary-tested safe list by Pandapi.

More from Pandapi 🐼

Pandapi makes friendly, no-nonsense data tools: simple input, clean output, and fair pay-per-result prices. Found a bug or need a feature? Open an issue on the Issues tab and we answer fast.