Download: https://files.catbox.moe/h4b8ve.zip (v2.3 now, small changes)
I've worked on it a little, and had some testing with it (on YouTube). I've added a lot of words into the blacklist, which currently is this:
"fag", "cunt", "dick", "idiot", "shit", "cuck", "simp", "fuck", "kill", "ass", "censor", "dissi", "jew", "yenta", "bank", "idf", "nigg", "coon", "negro", "black", "spic", "gay", "homo", "degen", "suicide", "tranny", "troon", "jeet", "brown", "retard", "hitler", "nazi", "natsoc", "goy", "communist", "mongrel", "race", "mixer", "rape", "groom", "sex", "genocid", "holo", "white", "atrocit", "subh", "media", "death", "murder", "extinct", "porn", "thot", "whore", "bitch", "foid", "massac"
Note: Words that contain any of those get fully obfuscated.
I had moderate success, so clearly the comments are not outright censored, but one of my comments using it was. I am not sure why - it could literally be due to its length, but bizarrely it went through (1 out of 4) when I obfuscated the entire text (all letters). I checked with an incogni window. One version was just the first 2 paragraphs, which didn't make it either. I'll write the comment below.
Other comments that used it normally with blacklist mode on went through.
So it's hard to figure out if the letters are actually parsed or not. But if they allow the entire text to be obfuscated, then just deactivating Blacklist mode might do it (it will obfuscate every vowel + a few more letters). Anyway, there are plenty of options to try around. I'll give an update when I have more certainty.
It could be semi-random as well, making it harder to assess their algorithms. Plus sometimes the comments appear with a 10-20 minute delay.
Summary of its purpose: The Obfuscator replaces letters with similar/identically looking letters that maintain readability to humans, but *may* elude censorship algorithms. Meaning you could write "FΑԌԌΟT" (this is the obfuscated result) instead of "F4GG0T" or with stars, and the censorship algorithm sees it as F????T where ? are *presumably* unknown characters.
Holocaust, Holodomor are likely both in their keyword list. You don't need to use the tool when talking about mundane topics.
> “White,” “black,” and “brown” similarly can’t possibly be shadowbanned
Someone else mentioned it - it's about the context. Or rather gathering various keywords. Take "black" and have other keywords like "crime", "statistics", "horrible", "mongrel", etc. "Brown" is meant for brownoid, which I like to use. "White" may cause the comment to be still flagged political, especially if capitalized.
> “Media” and “extinct” have to also be contextual.
I had trouble getting the simple message through (without the tool) that jews own 96% of media. Combine "jews", "media", "own" or "run", and that may be enough.
> What happens when the system starts shadowbanning all “unknown” characters, though?
That requires them to react to the tool, which will only be known if used sufficiently. That one comment I mentioned that consisted purely of weird characters went through apparently, so it doesn't seem to have happened yet. The exploit will only be known and addressed if it's prevalent enough.
But they can't really ban unknown characters though... they have to enable all characters for all languages. They have a built-in comment translator.
I guarantee it. But they can’t *just* block any use of ‘holo’ alone because that would also block ‘innocuous’ things. Like Holocene and holomorphic.
>Take "black" and have other keywords
>Combine "jews", "media", "own" or "run", and that may be enough.
Which tells us that their systems are annoyingly sophisticated...
It doesn't have to be sophisticated either, just err on the side of censorship.