YouTube is steadily incorporating more and more AI into its foundation. Not only did we allow Google to talk to Gemini about YouTube videos, but we’ve also seen reports that YouTube is using AI to “improve” users’ videos without public consent. But now YouTube seems to be taking excerpts from Sora’s book and allowing you to create your own AI avatar, which you can then throw into your video and make it part of the scene.
Now you finally know how to hide (and get back) your YouTube Shorts
YouTube doesn’t allow you to easily remove Shorts from your feed, but there are ways to do it.
Post your face in a YouTube Short without actually being there
AI picks up a mask in your shape
It hasn’t been that long since we said goodbye to Sora, OpenAI’s AI video app that lets you project your face and voice onto videos with a simple prompt. There’s still debate as to why OpenAI dismantled Sora, but the truth is that Sora’s death left a huge void in the AI video space that Google is now looking to fill with YouTube Shorts.
According to YouTube’s help page, the idea is to allow YouTube creators to record live video selfies of themselves talking to the camera. This video is used to create an AI avatar for that person, which can be added to various videos and scenarios using prompts to the AI. This is literally exactly what Sora was doing before, but this time it leverages Google’s Veo video model.
This feature is already rolling out to YouTubers 18 and older outside of Europe, and you can find out if it’s available in the main YouTube app or in YouTube Create. first,[リミックス]find the menu,[リミックス]Select. Reconsider > Add me to this scene,or Video creation options and select create a video with my avatar.
Videos will include an AI watermarkBut it’s an incomplete solution for calling AI
Like many of Google’s AI products, the new Shorts AI avatar system provides videos with visible watermarks as well as digital labels such as C2PA and SynthID. While technically good at identifying AI videos, physical watermarks added to videos are not a good indicator of accuracy in the world of social media. In the world of social media, it’s easy to crop or edit away watermarks.
Of course, this is where digital labels come into play. But the problem with SynthID and C2PA is that you have to bother checking if something is AI. People aren’t likely to do it that often, especially if they’re just scrolling through YouTube Shorts during a break from work.
Considering how big of an issue deepfakes are on YouTube, I doubt this will minimize the problem. However, if you’d like to try it out for yourself, it should be available on YouTube within the next few days or weeks.
