This must consume a tremendous amount of processing to do since they would have to transcode copies for every ad region/campaign and every resolution on demand. I am interested in how they made this financially viable.
Not necessarily. They could split the video in advance, assuming the ads will always be at the same point. Even if not, they could still use the direct, unaltered source with a range. The big challenge would be keeping it all synced, which I think is safe to say that they will get right.
But even if it did need to be transcoded, YouTube automatically transcodes every single video uploaded, multiple times. They are clearly not afraid of it.
No, they’re not split. Each one of those results you get from yt-dlp is a different version of the same video. I.e. different resolutions, different codecs. Some of them are the audio, some of them are the video, but they’re not split.
They don’t need to do any extra transcoding. It’s not that costly to stitch videos together. If done at specific strategic locations, it’s like copying a text file into another.
This must consume a tremendous amount of processing to do since they would have to transcode copies for every ad region/campaign and every resolution on demand. I am interested in how they made this financially viable.
Not necessarily. They could split the video in advance, assuming the ads will always be at the same point. Even if not, they could still use the direct, unaltered source with a range. The big challenge would be keeping it all synced, which I think is safe to say that they will get right.
But even if it did need to be transcoded, YouTube automatically transcodes every single video uploaded, multiple times. They are clearly not afraid of it.
If you’ve ever used yt-dl, you’ll know that YT vids are all split into multiple files. Presumably, this is where the ads get injected.
No, they’re not split. Each one of those results you get from yt-dlp is a different version of the same video. I.e. different resolutions, different codecs. Some of them are the audio, some of them are the video, but they’re not split.
They are also that. But when you watch YouTube-dl download a video, it downloads several parts, then ffmpeg recombines them into a single output file.
I’d guess its a solution similar to DASH that dynamically streams different content.
Since AI has caused them to complete abandoned any illusions about their carbon zero footprint I think they just stopped caring.
They don’t need to do any extra transcoding. It’s not that costly to stitch videos together. If done at specific strategic locations, it’s like copying a text file into another.
If they only do it for very popular videos, the additional cost will be trivial.