Lex Fridman: preference

Transcript context

…CineForm, nice. … yeah, at the time before actually led to the open sourcing of that work. So in parallel to doing the binary side, you obviously have samples. In many cases, you don’t have many samples so you have to figure out what all the different flavors are, and you may have a So CineForm, for example, is actually a collection of different approaches and toolkits within that codec ’cause often it grows naturally. And the hard part is finding a sample that gets you kind of somewhere to start without having to implement 10 different other things. So start there. I think thankfully at the time I found a sample by pure chance that had a lot of flat blocks. It was animation, so that really helped a lot because it wasn’t using particularly complex coding tools, et cetera, and you could kind of get somewhere and then, and then build up and build up until you figure, “Hey, here’s a few bits here. I missed this. I missed this, this if branch that it does,” and go, “Oh.” So when we say samples, you mean sample videos- … and then, and then you’re tracking, trying to infer, like, what is this codec doing- … by observing the sample and then looking at what, at the machine lo- The machine code saying-…

Stored transcript either side of the excerpt. The highlighted words are the published quote; the surrounding text is unedited source, never generated.