Lex Fridman
👤 SpeakerVoice Profile Active
This person's voice can be automatically recognized across podcast episodes using AI voice matching.
Appearances Over Time
Podcast Appearances
We can lower the frame rate.
We can use different kinds of codecs, as we mentioned, from H.264 to AV1.
There's ways to tune the trade-off between bitrate and quality, as we've kind of spoken to.
You know, you could do constant bitrate, you can do constant quality, say, or FCQ, QP.
You can do the longer or shorter group of pictures, GOP, that we mentioned.
I mean, all that kind of stuff.
It's crazy.
Number of B-frames.
Yeah, what is crazy is that...
Yeah, they're wizards.
Hats off to them.
YouTube, to deliver, all the streaming sites actually, to deliver at scale.
And YouTube is really magical because it's not just doing what Netflix does, which is one way, like broadcasting type thing.
It also has to upload videos from all the places.
So they're also doing encoding at scale.
For videos, they're going to be watched by like five people.
And it still has to deliver them at a moment's notice.
No delay, nothing.
Very minimal latency.
And also serve it in all different resolutions.