I looked into older topics but this issue is in the last one
- Device model: iPad Pro 3rd gen, iPhone 16PRO*:
- Version of operating system (IpadOS 18.6, iOS 18.5):
- Version of djay (5.4.1 / 5.4.1):
- Hardware/controllers used (just software):
When I started using DJAY, Neural MIX was working great but overtime it started to fail. I don’t know if it’s a software update issue or I started to use more and more and noticing inconsistences. For exemple, I will upload a video with a example of bad sound separation.
Sorry, but what do you mean? I can’t hear anything unusual. 
1 Like
When I turn off the voice knob on the last 4 bars it cuts more than that.
Hi @Gabi0110, sorry for the delayed response - this one must have slipped past me. I watched your video as well and I don’t see or hear the issue either. It looks and sounds like only the vocals are being removed to me. Could you please elaborate or provide another example? Thanks!
Yes, I will elaborate.
In my video, you can see a 16 bar loop and in the last 4 bar’s are with some-sort of vocals. In the beginning of video you can hear it but after I turn down the vocal knob, on the last 4 bar’s the sound will change instead of remain as the first 12 bar’s.
I’m expecting from DJAY to isolate only the “vocals” from last 4bar and leave the drums and harmony like in the fist 12’s bar.
I don’t believe that I’m the only one that notice the difference in the loop after I turn down the vocals 
Thanks for elaborating @Gabi0110. I’ve passed your video and explanation onto our engineering team for further review.
In the meantime, can you please navigate to the djay Settings>Advanced and confirm what your NEURAL MIX Quality is?
Sure, I also checked before posting. It’s 100% for both.
1 Like
Thanks for confirming @Gabi0110. Maybe try switching to 4 channel NM volumes and removing the vocal that way?
Nope, it doesn’t make any difference.
I noticed in other songs that the separation it’s made better but still the instrumental it’s a bit affected when acapella is muted.
Thanks for the follow up @Gabi0110. I discussed this with engineering and they explained this is not a bug. It’s simply a limitation of the current realtime STEMS separation technology. Please note this is not a perfect tool that will provide the same results as an original studio separation. Certain tracks are simply more difficult to cleanly separate using this technology. I hope that helps explain what you’re observing. Thanks!
1 Like
Got it, thanks for clarifying!
You’re welcome @Jaxon-Rhys. Also, welcome to the community!
