How many custom sounds do you make compared to just using sound libraries?

I'll leave that one to Phil, actually.

Well, pretty much all of them. Very rarely would we ever drag and drop from a sound effects library. Typically what we do, at a very basic sound design level: if you want to design something, say like an ambience or something, we bring in lots of different source material and then track-layer and make sure it supports all the different frequency content bands, like lows, mids and highs and stuff. And then process it and render that out as a new file. So it's not like we'll get wind and then just cut that out and drop it in the game. We'd always take it into a DAW and combine it with other different sounds.

What's a DAW, for most people?

A DAW is a digital audio workstation. Like Reaper, Pro Tools, Cubase, Logic, all that kind of stuff. So yes, we have access to loads of sound libraries, but we purely use that as source rather than end-user material that we push out to the game.

We do a lot of our own recording of our own source as well, hence the freeform jazz metallic madness. Because I think that's important, to get source that has unique provenance, when you know you've got something that no one else has experienced. It's a combination of both things.

You get a high burn rating on some sounds. Everyone knows the Wilhelm scream, right? When we hear that, you just cringe. I'll be watching TV shows, films, and there will be certain sound effects that you just cannot use, not even as source, like servo stuff, some pressurized air stuff, there's a flame whip that's overused all the time. And even if people aren't aware that they're hearing this repetitive material, people will subconsciously know that it sounds like something else. So we always try and move away from that and try and record as much as we can.

Yeah, the brain is a lot less forgiving to repetition than the eyes. Interestingly, I think there was a question I saw about how do you ensure that wild lines and stuff aren't so repetitive. A lot of it, we're looking at ways to use, say, stuff like machine learning perhaps, to process dialogue lines so that we can almost procedurally do variations rather than have to go back to the source and back to the actor. And that's very experimental stuff that we're doing with that. And that's a long way off. But it's in our minds, that sort of thing; you've got to keep bearing it up.