[19:11 Mon,3.February 2025 by Thomas Richter] |
YouTube is experiencing a flood of AI-generated faceless clips that steal content from "real" channels to make a quick buck. Successful clips produced by human YouTubers are automatically processed by AI into a new clip. AI summarizes the transcript of a specific video and creates a new voiceover text from it, which is then spoken by an AI voice and illustrated with AI-generated/animated images or a compilation of stock footage. Getting Rich Quick via YouTube?These clips, produced quickly and – partly with the help of appropriate tools – fully automatically and in large quantities, are then in direct competition with the original clips on YouTube. They are (still) recognizable by the fact that the narrator is not visible in the clip itself, that the voice sometimes makes strange mistakes (glitches), and that the images are very AI-generated or consist of more or less suitable stock footage clips.
![]() Numerous ![]() ![]() The Creators& Fight Against CopycatsReal YouTube creators are threatened with being overwhelmed by the increasing competition from such AI clips – they simply cannot produce as quickly as these. And they are naturally angry when their painstakingly created clips serve as "fodder" for AI tools to generate based clips – or even blog articles. The tech YouTuber ![]() Subtitles are "poisoned"This is exactly where F4mi comes in – her idea is to smuggle nonsense into the subtitles in addition to the real text, which is invisible to humans but cannot be recognized by AIs and thus renders their summaries unusable or "poisons" them. She uses the actually outdated subtitle format ".ass", which gives her the possibility to add styling instructions à la CSS to the text, such as font color and style or text position. With the help of these styling commands, she makes the new nonsense text invisible by placing it outside the visible area and setting the font size to 0. ![]() And indeed, AIs that are supposed to create a summary of such a supplemented video are so confused by the additional text that their output no longer has anything to do with the actual video and thus is not suitable as a basis for a "faceless video" copy. However, more intelligent AIs can recognize the trick – but F4mi already has ideas on how to blind even these – and even for the case that the spoken text is newly transcribed externally, the resourceful YouTuber has already devised a countermeasure. ![]() deutsche Version dieser Seite: YouTuberin "vergiftet" ihre Videos gegen Content Klau per KI |
![]() |