I get it. Technology is complicated. Press releases are long and filled with marketing speak that doesn’t exactly make sense. And new things are really confusing. I, just like the rest of you, want to be sure my purchasing decisions are driven by accurate information, and the internet is filled with the exact opposite. It is rife with confusion, misunderstandings, and what sometimes just sounds like willing ignorance in the face of something new and/or different.
After Stems got announced, a lot of websites were peppered with comments asking the same questions, as well as concerns and complaints about the Stems format that just didn’t make any sense. We here at DJWORX want to make sure you all make the right purchases and decisions based on as much information as possible. In that spirit I reached out to Native Instruments to get some clarification on Stems, the format, their integration and just generally how they work. If you have further questions put them in the comments and, if they are appropriate, I’ll pass them along and we will keep this discussion going.
One of the big questions that was being passed around that I want to get clarification on is how these files are packaged. I know there is a limit of four stems, but does the stem.mp4 file also include a full mastered version of the song, or is it summing the four tracks together every time I play the file?
The fifth file in the MP4 is the original master, that way you can still use the Stem file and play it on 2-channel systems like CDJs, iTunes, etc.
Related to that, if the completed, mastered file is included along with 4 stem tracks, is it possible for me to remove aspects of my completed song and only isolate pieces of it that I want in stems? For example, if I do not want to include an accappella stem, and only want to include four stems of instrumentals (drums, effects, lead and bass) can I do that, while still allowing the full file to play?
When playing Stems, the original master is not used at all. Only the 4 Stem tracks will be used. With all Stem volumes at max, you get the mix as intended by the producer.
Because the Stem tracks are not related to the stereo mix track, it is possible to have these two things be completely different. You could, as you suggest, make a Stem File where the stereo track has vocals, but those vocals are nowhere to be found in the 4 Stem tracks. Therefore, when a user plays the Stems version, they’ll never get access to the vocals. So while this is technically possible to do, I imagine it will seriously piss off customers who buy the content if they don’t get the same parts in the Stems as in the original master. It could turn customers off from certain artists, producers, or labels if this tactic was used often.
So, you either play the stereo master without Stem control, OR you play the Stems. There is no in between. This is not some kind of hybrid where you can play the stereo master and then somehow switch to the stems right in the middle of playback, or use the stems to somehow “subtract” content from the stereo master. You’ll have to make your choice about which content to use from the start as there will be noticeable level shifts, timing shifts, and other anomalies if you tried to switch from one source to the other in the middle of playback.
Is there a way for me, as the user, to extract the stems when this format is released, without using Traktor? If I use Serato, will I be able to split this file apart and access the stems in any way?
Yes, because the Stem file is an industry-standard MP4 file, not something weird and proprietary. You can get a Stem File and load it into Audacity and access the 5 tracks it contains. We will publish all the specifications for the Stem File so that anyone—even Serato—can implement this. If they do, then you will be able to access the individual stems from that software. DAW manufacturers could offer direct Stem import and export functions if they wish since we will publish all the data on how to create these files and even provide the compressor/limiter. This file format is truly open.
Can you provide any updates as to other applications which will be supporting the Stems format on release
besides Traktor?
None yet since we have not yet published the specs on this format. But as soon as we do, any other manufacturer / software company can integrate this stuff. Reading the file is actually pretty easy—we really tried to keep things simple so everyone could make use of it.
This obviously isn’t every question you guys have, but it was what I could glean from the various complaints and concerns I saw passed around the internet. This new file format isn’t a world changing, brand new concept, but it’s the first time a major developer has proposed a viable standard, let alone given it away for free. It might not change anything, and it might not get adopted, but there are very few instances where new technology and standards have been bad.
As I mentioned elsewhere, the 4-stem version and the full mix down are going to sound very different. Not necesseraly a bad thing given the intended use but something that might surprise a lot of people.
That is probably the intention. I mean, if the producer wants to take the time, they COULD make them fit almost identically when the stems are played together, but I don’t think that’s nearly as exciting.
It’s impossible. You can’t sum 4 tracks, then master or master 4 tracks, then sum and have it sound the same. Mastering takes into account and reacts to all the little interactions between the different parts.
The resulting 4 parts will either sound different when summed or sound different when isolated. There’s no way around it.
My guess is that people will go for the ‘sounding different when summed’ solution as most users will want to isolate parts.
Not really a criticism though, it will work well for what it’s intended for.
Makes sense. It would need extra work, hence the monetization factor.
Let’s not forget that the original tracks are likely to have considerably more than four stems. Summing the supplied stems is still going to be missing a lot of the original track.
yeah that’s a good point. I was going under the assumption that most producers would make the stems reproduce the full track but that’s totally not a requirement.
Some people might only put elements of the track in the stems.
Yeah, it’s almost as if this was intended to cause confusion.
Yeah, this concerns me. Noting that even the F1 has 16 pads, why would they only release 4 stems per song? Why not include all of the stems (or at least up to 16)?
Ok, that gripe aside… when you have stems (aka individual sound sources), you’ve crossed the event horizon to have the pieces you need for spatial sound. At this point, you can move sound around in space, and (if you’re able to save transforms over time) animate them. Will NI be working on a spatial sound plug in that allows for this kind of usage?
On another thread, I’ve suggested resurrecting Kore for use with Traktor and Maschine, and furthering this integration, enabling 4 speaker surround sound. How will Maschine implement Stems?
I think the 4 stems thing is because it is less. The nice thing about this being open source is if there is a demand to use 16 different stems. That would mean breaking out 16 different elements of your song. I’m not sure how much demand there would be for a lot of those elements separated out from the main mix.
The implementation of this wouldn’t be to have a Kick drum track. While that’s great for a lot of us, I think the idea is “Here’s the drums” and “Here’s the synths” etc. Keeping it like that can allow the producer to provide more tools for DJs without compromising their signature or unique sound.
Keep in mind that the only thing that NI is really ‘open sourcing’ here will be the format of the MP4 atoms for the stems names and their colors. Everything else is just them using the standard MP4 container format, which has no limitation to 4 streams or anything like that.
So it’s all down to the software using this, Ableton could make a version outputting 16 stems if they wanted, just as an interchange format for example.
Precisely. NI has set it up with 4 stems, but anyone could change it to however many they want. I have a feeling there’s a point where there are diminishing returns, either for the producer OR the consumer. I like the idea of 4 well curated stems, put together by the producer. 16 could be great, but I’d rather take the 4 and break them up into loops and samples to create 16 cells.
I agree. I think the current number 4 comes from the number of faders in a typical mixer or on NI’s remix decks.
If you don’t have the hardware to controls the gain/volume on each separate stem then this becomes a lot less useful.
Right, exactly. But, as that demand changes, it becomes a lot more interesting. For example, using this with a Remix Deck is a LOT different than using it with a full deck, and can offer more possibilities. But even just loading stems into something like Audacity and making edits to them is really exciting.
Here’s an inherent problem with playback in this manner: sound placement. If a sound is played dead center, it lacks any width. This is why Producers place (and sometimes move) sounds to the left, center, and right in the stereo field. If you play all your sounds dead center, it sounds… just like AM radio mono. Which is to say, it sounds like it’s coming down a hallway.
Stems are stereo tracks. There’s no reason for them to be dead center unless the producer wants them that way (for a bass or kick drum stem for example).
Ok, then this is slightly confusing, I know Traktor is designed to output stereo even if only one Main is used for a mono event space.
Native Instruments’ Stems have a slightly different meaning if these stems are in mixed down Stereo (with no pan control). I’m familiar with the industry standard most Producers use recording live artists at the source of the sound, not the room, specifically so the Producer can mix each piece better (most typically for balance in the mix, but also for showcasing a soloist When they play).
There is a valid argument to provide each stem in mono, and have a default pan position on each stem voice, as this would give the Producer what they need, and at least start where the Artist intended.
This is at least my take on the issue. I do hope the tracks are recorded in mono, and presented in Stereo.
I’m not sure I understand want you are saying but all I can tell you is that stems, like remix deck samples, are stereo tracks.
What goes into those stereo tracks is up to whoever produces the content. Typically this would be coming from a DAW session of some sort, with panning and widening already applied (if any).
Yeah, I think that’s the difference between remix elements and stems. So maybe we need better nomenclature here. I’ve always understood clean samples of tracks (like the Amen break), centered, with no effects on them to be clean stems. These can be used anywhere and mixed with ease. When I’ve received tracks pre-mixed, with some effects applied, I’ve considered them to be a remix kit.
I’m totally open for others to chime in, but these are two different things.
I think there is a big difference between traditional stems and this new format from NI.
As a producer, I would love to get 16 dry stems to work with. But as a DJ I think NI hits the sweet spot with 4 wet ‘stems’. You will be mixing at least two tracks so there’s 8 stems to work with right from the start. Enough to be creative but not too much to make you lose momentum or presence.
While that is true, the F1 and the S8 still have 4 banks of 4 (only one from each column at a time). You still only have 4 channel strips, but you have 4 choices per column instead of one choice (which is really the choice to have each element on or off). I guess NI will be attempting to create “section mixing”, and I guess this could be done via the deck assignment to crossfader option (Maschine has a behavior opposite of this right now, choke groups), however I’m not so sure this was planned from the beginning for Remix Stems.
Instead of talking about it, I’d prefer to help as part of the Dev Team. I hope someone at NI is reading this.
I’d really like to have clean (dry) vocals, because sometimes, I hate the vocal effects, and sometimes I just want to change how the lyrics are used.
That’d be cool, but it’d be completely up to the producer. I can’t imagine that being something we see a lot of, unfortunately.
Yeah, but there are different parts for each instrument in each song, and there are often more than 4 instruments in a song (including vocals). Maybe this is the stems they want to sell us, but not the stems we want to buy? I used to do a lot of studio work (and still do from time to time), so this begs the question (for me): will this be laid out like a 4 track recorder (with pieces played solo over time), or will each instrument have a discrete, sectionalized, group of all elements for each instrument?
It will be whatever the label/producer decides to do.
Native said they will release a ‘best practice’ document that will encourage people to standardize this but everyone will be free to do what they want.
In the current examples of stem files available, some of the 4 stems are things like ‘drums’, ‘bass’, ‘synths’, ‘vox’ while some others are ‘kick’, ‘clap’, ‘hi-hat’ and ‘dub’.
There’s also the possibility for different stems in different tracks.
Well, whatever happens, we’ve crossed the event threshold for spatial sound. Moving sound in an environment requires individual sources that can be manipulated… this is that.
I’m more than ready for this, I’ve been waiting nearly 20 years for it to happen.
I think it’s going to work however the mastering engineer/producer wants it to, but it will be four audio files. You aren’t going to be able to break the drum out automatically to each individual loop involved in the track, but you could do that in Traktor manually by taking loops and dragging them to Remix Decks (in theory). Since you can break these out in something like Audacity it gives you a bit of control, but I have a feeling it’s going to exist like a capellas, except for whatever the producer wants them to be, however they want them mastered.
I’m almost 100% sure, though, that this is just going to be four tracks of audio. That’s it.
And if it’s only “Stems light”, someone else will have an opportunity to do what Facebook did to MySpace… offer more of what the people want despite being second to market.
It concerns me that even though they don’t want to control what people do and don’t include, they should have established some guidelines and best practices procedures.
I think I’ve mentioned that somewhere else but it says so right in the stems FAQ :)
“Are there any best practices on how to create a Stem file?
A document outlining best practices for instrument grouping, order, names, colors, and mastering techniques will be available from the official Stems website.”
I believe I missed that. Thanks for the correction.
They… are establishing best practice guidelines. That’s in the initial story, as well as the one we’re commenting on now. And it’s open source. It’s not THEIR product. They are releasing it to the world and if someone wants to build a package for 16 actual unmastered stems then they can.
And, at that, you CAN put in stems like you’re talking about. It’s just not the intent. This isn’t for remixes, this is for live performance.
And along comes Dubseed…
I think the idea is that the full track is going to be head if all 4 stems are playing. So there will need to be a mixdown of all the different traditional stems in a production into the four ‘stems’ of NI’s format.
It’s already discussed elsewhere but yes, there will be a mixed down version in the file. That version will not be accessible if the Traktor deck is set to ‘stems deck’ only if it’s set to ‘track deck’.
You absolutely can make a mastered stereo track sound like a four-stem collection of tracks played together. Strictly speaking, mastering is a stereo process that happens after final mixdown. Applying a given set of stereo processing steps to a single stereo track or to four stereo tracks and then summing them should yield the same end result, assuming that the recording and mixdown engineers did their job vis-a-vis phasing issues.
In the real world, lazy producers will likely continue to rely on sending their multitrack projects to engineers to fix the problems with the track and call this process “mastering,” so it may be rare to find a stem file that accomplishes this feat.
I’m not sure you understand how this works. Let’s take a simple example: compression.
You put a compressor on your master chain. At one point in the track, all the 4 parts summed cause the compressor to compress the master by, let’s say, 3db. You store this result as your mix down. Now you take your 4 parts and apply the same compressor on each separately. At the same point in the track, none of the 4 parts are loud enough onto make the compressor work, no compression occurs. If you sum the result of the 4 parts mastered separately, it will not be the same as the full mix down because one is compressed and the other isn’t.
Now you could approach this another way and try to split the stems so that when they are summed you will always get the same signal as the mix down. If you do this, your stems will not sound the same as the 4 source parts. Given my example above, the 3db compression will have to be split across the 4 stems or applied to one of the stems, either way it won’t be the same.
Compression is just one thing but if you add things like distortion or saturation or limiting, things get even harder.
I hope this makes sense. I’m not trying to be a smart ass but this is really how things work.
It could be said though, that putting a compressor (etc) across your entire mix is bad form. If you’re producing with multiple tracks, then you should really be processing each track individually as required, to reach your whole. If you’ve done that properly, then any “mastering” required should be minimal.
I’m not sure where that could be said :)
Every master chain I’ve seen before always has at least one or sometimes multiple compressors which are used to ‘glue’ the parts together (one, by cytomic, is actually called the glue just for that reason).
But hey, I’m not a pro-producer so I’m just going to respectfully agree to disagree on this.
You can use sidechain to have the compressor on the each 4 stems react to what happens on the proper stereo master.
You can, but then you’ll have the stems ‘duck’ in some places when they are soloed which might not be what you want.
That’s what I mean by “you can master the stems for the summed version to sound right or master for the individual stems to sound right”. And I think most people will go for the individual stems since that will be the typical use case.
These are guidelines, I expect all kinds of variations as that’s how messy things are in real life.
Since this is open source, it seems like Ableton Live and Bitwig will be the big beneficiaries of this format more so than Traktor or other DVS setups.
If this new format indeed takes off down the road, music libraries are about to quadruple in size. A 1TB drive to accommodate you music library won’t sound so huge.
At least storage is getting cheaper.
Well if it takes of I only se myself getting a few tunes in Stems format, which will be the ones I like to use the most when doing live mashup/remixing. Like I have it now with a few acapella files and drum beat files. So for me it will only be a little increase
I was looking at 6TB Hitachi drives the other day for my NAS, quite pricey but thats a fair bit of disk to fill.
Nice article. Thank you
the one thing that worries me about this is that DJs will be charged more for these stem files than for the original tracks. If DJ pools offered them as part of their package then great. I doubt everyone is going to want to use them and maybe only on a selection of tracks. Which may limit their uptake somewhat. It is like having the Really like where this is going otherwise increases the creativity options.
Storage is not a concern really unless you are a mobile jock who needs to keep everything and then you are not going to have STem versions of everything.
Keeping your collection lean is the hardest part of being a pro DJS these days.
To take the opposite argument: Producers are basically giving away the recipe to their secret sauce with these files, and as such, deserve to be recompensed for it.
To add more to that, the record companies will up the price and the producers not see a penny more which usually happens. Unfortunately the music industry is like that. Squeeze the provider and the consumer at both ends of the deal.
I’d still like to know whether we can do this stem file business with lossless audio quality or not? To me that’s going the big one.
As it stands currently it’s AAC compressed audio only.
I don’t see why FLAC couldn’t be altered to work the same way. It’s just a container file, after all…
They actually named both the compression and the container format FLAC just to be confusing enough :)
But, yeah, there’s no reason why it couldn’t be done as part of FLAC as long as the container supports multiple streams.
Stems are perfect for my original performances. Being able to mute layers depending on what musician joins me on stage is a godsend. At the moment I have to have multiple versions of each song to cover every possible combination. But just having one version would be awesome :)
Though I doubt I would actually DJ generally with stems.