Image Recognition could attatch the first frame of an ad to the length of time the ad plays for, then add it to an online DB a la sponserblock.
They might try to block seeking during these sections, but YouTube usually has raw mp4 streams available under the hood. You can even pull them using invidious or newpipe. Take that out and we might be fucked.
A good way to block this kind of thing is just to use DRM. Most platforms now provide a completely blocked off and secure hardware DRM solution that makes it impossible to grab video frames or view decrypted data in any way from the host operating system or any app running on it.
Ripping the video segments would just give you encrypted and useless data without the license.
These kinds of systems would need to be attacked by HDMI or other downstream hacks, or an HD video camera pointed at the screen in a dark room :)
I don’t know youd be able to do it within a browser extension but something like newpipe or yt-dlp?
Public Invidious Instances would be tough because that’s a lot of load to stick on a server, especially one run by a hobbiest. But self-hosted single/low user instances could also feasibly do this.
Obviously its gonna take a good bit of work, but it IS doable.
Image Recognition could attatch the first frame of an ad to the length of time the ad plays for, then add it to an online DB a la sponserblock.
They might try to block seeking during these sections, but YouTube usually has raw mp4 streams available under the hood. You can even pull them using invidious or newpipe. Take that out and we might be fucked.
A good way to block this kind of thing is just to use DRM. Most platforms now provide a completely blocked off and secure hardware DRM solution that makes it impossible to grab video frames or view decrypted data in any way from the host operating system or any app running on it.
Ripping the video segments would just give you encrypted and useless data without the license.
These kinds of systems would need to be attacked by HDMI or other downstream hacks, or an HD video camera pointed at the screen in a dark room :)
Its bad enough they use widevine on their free movies/shows but the idea of them requiring widevine for regular YouTube sounds awful.
Hopefully legacy clients/devices will stave that off until something else comes along.
Show me the image recognition API in the browser docs :)
I don’t know youd be able to do it within a browser extension but something like newpipe or yt-dlp?
Public Invidious Instances would be tough because that’s a lot of load to stick on a server, especially one run by a hobbiest. But self-hosted single/low user instances could also feasibly do this.
Obviously its gonna take a good bit of work, but it IS doable.