AI Text: How to Make it Sound Human (and Avoid Detection)

Truth be told, texts generated by language models often feel cold and predictable. Sentences are too perfect, expressions a bit generic, and it's immediately clear a human didn't write them. It's a bit like reading an instruction manual: clear, but soulless.

This issue has become increasingly prominent. Schools, universities, and companies are all trying to sniff out machine-produced content. Using raw AI text can be costly, both in terms of credibility and evaluation. Not ideal, is it?

The solution comes from an open-source project called asokurasu/text-humanizer. This tool takes your AI-generated text and runs it through another language model, but with instructions to make it sound more natural. The asokurasu/text-humanizer project is available on GitHub and uses a multilingual LLM for its rewriting pipeline.

It's like having a personal text editor who never takes vacation and works for free. The model tries to vary sentence structure, add minor imperfections, or inject a more conversational tone. The goal is to mask that "robotic" sheen AI detectors are so good at spotting. Pretty clever, right?

The big selling point is that it's completely free and open source. You don't have to shell out cash for monthly subscriptions to similar services, which often cost an arm and a leg. Anyone can download the code and use it, or even improve upon it.

What's more, it's multilingual, meaning it isn't limited to English or Italian. If you write in multiple languages, you have one tool for all your needs. Asokurasu's asokurasu/text-humanizer project was published on GitHub, making this technology freely accessible to everyone.