The track is controlled by the deck, that includes the pitch and speed etc. The pads isolate the parts you want play with, ie drums and/or vocals etc.
3 Likes
Well, since I started this topic I enjoyed reading all of your comments, from the first where nobody was confident that one day we would have STEMS everywhere… Now the topic has grown so much and if you want to make it a community, well it’s OK for me… Anyway I’d like to thanks everyone who has contributed to this discussion, even the most skeptical, at least at the beginning… Remember, you’ll never know what the future will bring us, especially in this DJ world… That said, have a good time with your STEMS! CIAO!
With the pitch-fader?
2 Likes
Look at how stems work in VDJ, and especially with the new stems swap sampler. A multitude of different ways to be more creative. I bought a FLX10 with dedicated stems support to take full advantage and it has really transformed my mixing capabilities.
The Denon implementation is barebones at best.
- Optimistic reply: Barebones…at the moment. Version one.
- Pessimistic reply: Stems will probably go the way of the sampler. No enhancements.
1 Like
I guess it’s the difference between a privately owned small company with a dedicated dev / beta team and an owner who still actively codes and a global corporation with multiple brands and one shared dev team.
Indeed, Reminds me of all the hype behind the MPC Stem function, months down the road, no one is shouting about it anymore, Just commenting on how poor quality they are…
Also an InMusic company btw
Virtual Dj has been in the stems game for at least 6 years, they have a certain lead over the rest of the industry in this area. And the little guys at atomix have always been at the forefront of new features (the scrolling waveform is them, the beatgrid is still them etc…)
Denon DJ is just catching up on the stems features and even if the new algorithm is much better than the very first version presented over a year ago on the prime 4+, it still needs to be refined.
It is not bad it is just in the upper average compared to what other competitors are doing on the market. This is how I would currently rank the quality of Denon DJ’s stem rendering algorithm among the major players in the industry
Best stems algorithm:
- Virtual DJ Stems 2.0
- Djay Pro AI
Quite good:
- Serato Stems
- Denon Dj Stems
Average
- Tranktor Stems
- Rekordbox Stems
Denon Dj is currently positioned well above the Traktor Stems, but slightly below the Serato stems, which is not so bad for a first try. The quality of Serato stems already satisfies the majority of DJs who are the least demanding among us.
It should also be noted that this is a personal and average ranking because the quality of stem separations will depend enormously on the type of tracks and also on the way this track was mixed/mastered. This can also depend a lot on the musical genre.
Also keep in mind that many current productions, often Remixes/Edits/mashups are produced and mixed from tracks that have already been stemized previously (sometimes with a quality of stems used that is questionable and close to what we had in the first beta prime 4+).
But drowned in an edit or one to which you add instruments and therefore frequencies to mask the muddy sound of the stem used, it works. (Block & Crown are the specialists in this for example)
So if you re-stem a production that has already been created on a stem base, do not expect exceptional quality during your tests.
2 Likes
Absolutely not. The first stems release was June 2020. But still that’s nearly 5 years later for In-Music to get to roughly the same place on Denon gear.
Not to mention the quality of the original audio files! If the files are very compressed and of dubious origin (like a 128kbt file converted to 320 or other 92 to 320) in mono or false stereo you will not have a quality Stems. If the Stems are created from real Wav, Flac or AIFF files then I think that the rendering will already be much better!
3 Likes
Humm? 
Ok 
Let’s test that! 
I took a lossless version of Bohemian Rhapsody, and converted it to a 128kbps mp3 with File Converter for Windows.
Next, I stem processed both, then played them on my Prime 4.
You can download a short video here. It also includes the original wav recording from the Prime 4. This link will expire on 12/01/2025 at 18:52.
Do you consider this a big difference? I have to say, to my ears they sound pretty equal. I’m actually wondering how many people can point out which channel plays the lossy file, and which channel the lossless.
1 Like
Here’s the thing - stem separation is pretty miraculous tech. In the past I’m sure most of us would have thought it impossible. Here we are though, with quite astonishing results, considering the complexity of the task.
…and it’s that complexity, the sheer amount of processing that’s need to achieve it, which means the results are always going to be flawed. Even the most pristine recording is going to contain artifacts.
If you’ve got the ability to fine tune it to every individual case then it’s likely to be better, but a generic algorithm that has to deal with whatever’s thrown at it - it’s like the difference between a fixed focus camera and a high end DSLR.
5 Likes
Cher Believe does not play well with stems at all due to the synthesised vocals. But most tracks are fine.
Teddy Pendergrass - love TKO sounds awful too, but then no computer on earth is built to AI Teddy.
I find Belo Horizonte has the brass section in the vocal. I’ve found this with other saxophone & brass horn based tracks too. I think it’s getting confused with the vocals for sure.
I’m really happy with how it sounds for my usage and I’m sure it will be refined along the way. Just a tweak here and there.
1 Like
I tried The Jackson 5 - I Want You Back and found that the background vocals were missing from the “vocals” stem.
In general I find that with older stuff, the bass doesn’t get extracted - presumably because it’s either not deep enough or not forward enough in the mix.
Stems don’t like separated stereo as you’ve noticed. Some of the results for mid 60s stuff are way off.
Hello I had some time to test stems on mixstream pro go for riddim, grime including some rap genres and not everything was good. If you Turn on vocal stem only I still heard other layer of melody. With Enabled bass stem only, noticed high-end from Snare transient (very weird). For drums stem the drums based genres sounds too deep like it missing treble. Melody for me of other tracks has some other vocals layers in this.
I think the AI of stems should be upgraded in that case for desktop software.
I thought that “AI” for stems works for any mixes frequencies, masters and music genres but it’s not perfect.
Can’t wait for updates 