Hacker News Reader

From what Apple say about spatial audio, you cannot just run it thru a converter. You need the tracks in their individual original recording before the mixing to be able to do it properly. So I’m not sure how a handful of chop shops are doing all that.

View on HN

You don't "need" anything to make a track with spatial audio, it's just a format that you can package your sound in. You could feasibly turn a 48khz lossy MP3 into a spatial audio track, but it would probably sound pretty bad. Likewise, the actual stem-separation process that you're describing is similarly dubious: spatial audio can still sound pretty terrible, particularly in cases where it's an afterthought (see: all media).

Yeah you’d need the multitrack original recordings to do a good job I’d say. There are tricks with eg. phase cancellation/spectral filtering and so on I suppose you could use to try and hack something together, but the results would vary greatly depending on the original material/skill of the engineer… A lot of multi-track recordings, especially older ones, have a lot of bleed between tracks which might complicate matters…

Reply on news.ycombinator.com