Hacker Newsnew | past | comments | ask | show | jobs | submitlogin

Djay’s killer feature is algorithmic stem extraction. The streaming audio services that have integration (Tidal, now Apple Music) don’t allow this processing. It’s an arbitrary technical limitation that I suspect is rooted in a licensing agreement from the rights holders.

tl;dr this is great for playing with sequencing, but a lot of modern djays will want to be able to do stem separation even when exploring, and live use was always a tough sell with a no-internet kill switch.



edit: I see there's a tier that does allow stems again now! https://tidal.com/DJs

--

Licensing indeed. For a time VirtualDJ allowed stems for Tidal streamed tracks but were required to disable it.

https://www.virtualdj.com/forums/254722/VirtualDJ_Technical_...


Oh neat, thanks!

Yeah Virtual DJ integration was great, I got frustrated with their beat gridding (I come from Ableton where it’s always been magic and the Traktor where auto grid was really good) but they’ve been pushing everything else forward for a while.


Does it stem in real time? That is can you use it to do karaoke on an arbitrary song as it streams?


When you import a track to your collection, it does the processing. So, not completely in real time, but it only takes like a minute of background processing before it's ready. Virtually all DJ software has this as a feature nowadays, but of course, results vary.

It's like the next "big thing", very few DJ controllers on the market right now have dedicated buttons for it, but that's about to change.


With a GPU stems take seconds to generate, it happens on the fly when I load a file and the track can begin playing immediately while its running.


i've been using "AI/ML" to stem songs for something like 7 years, i was wondering if the state of the art allowed streaming stems yet or not.

"next big thing" heh


Stemming takes about 15 seconds for a 3 minute song. And most DJ software downloads the entire song in a few seconds, then generates the display waveforms, and so on, and then you can play the song.

So "realtime" and "streaming" don't really apply, because the files download and process fast enough.


15 seconds for 3 minutes is quite slow, the open source thing i use, spleeter, says 100x speedup over realtime with a GPU, that would be 1.8 seconds for three minutes, which means that spleeter could, in fact, given all of the arguments of what i asked, do it "realtime".


It depends on the hardware. The laptop I use for DJ doesn't have a dedicated GPU.

I also use open source, preferring demucs over spleeter.

In my experience spleeter requires huge amounts of memory, my DJ laptop only has 16GB of RAM, and I need to close the browser to be able to successfully stem a 5 minutes song.

Demucs takes a bit longer, but it uses just 2GB of RAM. And the sound quality of the stems separation is much better with demucs, which is a more important factor than the runtime, IMO.


thanks for the tip, i'll check out demucs!


It works on serato


I just tried it in Serato and it says stems are not available for streaming music. Their wiki says so also(0).

0: https://support.serato.com/hc/en-us/articles/12188545880719-...

Real bummer, I still have to pipe stuff through Ableton, record it, then import it..


Rats. I DJ’d a wedding with tidal last year and it worked. It just needed analysis.




Guidelines | FAQ | Lists | API | Security | Legal | Apply to YC | Contact

Search: