If you have recently been paying to YouTube, you may have noticed the growing trend of called “faceless youtube channels” That never presents a visible human who speaks in the video framework. While some of these channels are simply written by shy humans, many more are Totally automated through tools with AI To elaborate everything, from scripts and voices off to images and music. As expected, this is often sold as a form of People a quick dollar in the YouTube algorithm With a minimum human effort.
It is not difficult find Youtubers whining About an avalanche of these faceless channels Steal your integrated transcription files and Executing them through the summaries of AI to generate their own instant imitations. But a youtuber is Trying to defend himselfSowing their transcripts with junk data that are invisible to humans but poisonous for any AI that dares to try to work from a scalf -escalation file.
The power of .ass
Youtuber F4MIthat creates Some excellent deep immersions in Dark technologyrecently He detailed his efforts “To poison any summary of AI to try to steal my content to make a slope.” The key to the F4MI method is The subtitle format .s.Created decades ago as part of the Fansubbing Advanced Substation Alpha software. Unlike the simplest and popular subtitles formats, it admits elegant characteristics such as sources, colors, positioning, bold, italics, underlined and more.
It is these elegant characteristics that allow F4MI to hide garbage that confuses AI in its transcriptions of YouTube without affecting the subtitular experience for its human spectators. For each part of the real text in its subtitle file, it also inserted “two pieces of text outside the limits using the functioning function .ass, with its size and transparency established at zero so that they are completely invisible.”
In those “invisible” subtitles pictures, F4MI added text of public domain works (with certain words replaced by synonyms to avoid detection) or its own scripts generated by LLM full of completely invented facts. When these transcription files were fed in popular’s summary sites, that garbage text ended up overwhelming the real content, creating a totally unrelated script that would be useless for any faceless channel that tries to exploit it.