An audio format for creative DJing | Hacker News Reader

An audio format for creative DJing

stems-music.com

34 pointsahq23 comments

Why define a new AAC subvariant and not use an existing (and actually unencumbered) multitrack stem format like mogg that's been around for for a decade now?

I suppose there are more phones/mp3 players playing AAC files than the ogg format. It might gain more traction with this format. And the company behind it (Native Instruments) is really popular accross DJs and electro musicians (they do a lot of great hardwares and softwares).

So that the files are playable in any audio player that supports MP4.

Actually you mean exactly one player: iPod/iPhone. To play nicely with apple, you have to fork over the $2.50/copy. Ah well.

(Almost everything else currently using stems is mogg-based given the per-unit royalties for MP4 software support)

AAC is played by many devices, it's not an Apple standard, and the licensing fees are less than MP3.

Apple is one of the few vendors that supports AAC in exclusion of other formats (and, eg, ALAC in exclusion of FLAC, etc). And no, actually, AAC's declared licensing is considerably more expensive than mp3 (partly because AAC is a bundle of quite a few different codecs that share the same name).

According to that description, the format contains 4 separate tracks that together make up the song. How did you settle on 4 tracks?

Most tracks can be split up into four distinct components: drums, bass, melody, and vocals. Of course, they could have taken it a step further, by splitting a track up into its individual instruments, but I doubt that would be very usable in a live performance setting (although some performers actually do go this route, usually relying on software such as Ableton Live for this purpose).

The four channels are also quite similar to the EQ section of a DJ mixer which generally controls three frequency ranges (low, mid and high). EQ-ing is heavily used to mix songs together, so by sticking to a similar interface, DJ's can easily pick up this format.

I'm more of a "for fun" DJ (occasional parties, events, but mostly just my basement) but I'm tempted to try mapping the 4 stem tracks to the 4 knobs on my basic controller (gain, high, mid, low) in Traktor since it apparently supports the format. Probably not something I'd want to have permanently enabled but an easy way to play around with it. I guess if I wanted to map a key to toggle between EQ and stem track levels it might work.

I was wondering the same, and found that there are already available stems on Beatport, where they classify the 4 tracks as: Drum, Bass, Vocals, and Theme. https://pro.beatport.com/stems/about

I suppose DJ equipment manufacturers like Pioneer will step up and add volume controller, and mute and solo buttons for each track on CDJ units. So the fixed number of 4 tracks makes the hardware easier to design.

They already did it with remix packs, and all their gear is created for working with 4 tracks anyway.

1. Why oh why is 'there are 4 tracks' hard-coded into the format? What a perverse limitation. I understand there has to be a limit if you want hardware players to mix in real-time but 4 seems too low and Stem v2 will probably be needed next year to get round this.

2. This seems like a wasted opportunity in terms of adding extra features that wouldn't complicate implementation terribly much. Tracks could allow time offsets and repeats - you've then got a loop-friendly pseudo-tracker file format at very little extra cost.

Hi, my initial thought was the same. Why limit it to 4? Although I have recently learned the idea behind their decision was that most DJ music can be broken down into 4 elements. Drums, vocals, bass & riff. Having realised this, I now tend to agree that it should be limited to 4. Another reason I have heard was that they limited it to 4 due to Native instruments hardware limitations and playing several tracks at once would cause performance issues.

I have been producing electronic music for some time now. I don't think I have ever had a piece with less than a dozen tracks.

Yes, in the end it boils down to the 4 mentioned, but while you produce, you want way more than that.

It is mostly aimed to do lives, so 4 tracks is generally enough.

1. When you DJ, your main resource is attention — and while I have experience with Ableton live sets, trying to juggle more than 4 channels of a track that you sometimes heard once or twice before is not a very good idea, tbh. Also, Stems format was created by NI after they already succesfully creates Remix Packs format, and in my experience, 4 channels that Remix Packs have is quite a good balance — not to say that all NI Traktor gear is created for 4 channels in mind.

2. Once again, extra features not only complicate implementation, but also user experience. What exactly would you be able to do with time offsets and repeats that you can't do with Traktor's deck looping and loop recorder already?

> trying to juggle more than 4 channels of a track

sounds more like a tooling problem, stems should be able to be grouped by type with little effort (e.g., all the drumish things visually represented as one fake stem)

First of all — no, it's an attention problem. I've used APC40 in a live performance once, but in live perfomance (where you play your own music in more or less pre-determined fashion), unlike the DJ set (where you play other people's music, changing the playlist dynamically according to the crowd), it's easier to manage. Apart from monitoring the crowd, searching the next track, marking it and pre-listening (because unless it's in the hundred or two hundreds of tracks you know by heart, you may not know what's the best point cue it it and how), playing with FX, letting the photographer park his backpack under your table, answering to some random question by the manager, you also should have some amount of your attention on reserve, so that when some unforeseeable shit happens, you won't butcher up the mix and people stay entertained.

Regardless, if you already want to group the "drummish" stems, then why would you want to deal with the separately in the first place?

I get it that NI is mostly pushing this for the "creative DJ" market but I wish it was 8+ tracks and they got majors onboard to release stems of older music.

There are a bunch of leaked multitracks on the internet, everything from Metallica to Stevie Wonder and Michael Jackson - and they're really interesting to listen to and play with.

The original multitrack tapes are fast becoming degraded and lost so would be fantastic if there was commercial motive to archive/digitise them.

NI's products such as the S8, D2, F1 and X1 are mostly targeted at DJs playing electronic music, particularly house and techno. Genres where there's a lot of room for live experimentation with many different layers, due to the easily quantisable nature of the music. I doubt they had Stevie Wonder and Metallica in mind while developing this ;)

For production purposes you'd probably want a lossless file format anyway.

This looks awesome. Ever since I've been DJing, this separation of components is what I have been wanting to play with.

The 4-track limitation is stupid. Splitting a track into drums, bass, vocals, and melody instruments is definitely useful for a lot of purposes, but it's limiting for real in-depth remixing. Why not make it unlimited?

The 4-track limit means you can build a hardware DJ controller that supports stems. It also means your software doesn't have to have a crazy unlimited-track interface. 4 tracks is the sweetspot where you can actually build usable products around this for live performance.

Clarification: When I say build hardware, I mean design a usable hardware interface for mixing 4 stems. I don't have to guess what audio is on what track. Imagine what a nightmare it would be if every stem had a different number of tracks and you were trying to DJ live with them.

Reply on news.ycombinator.com