Here’s a little tip that I’ve been using for the past week:
It’s quite challenging to find good backing tracks that sound natural. Luckily, with many people dabbling into AI, there’s quite a lot of stem separation algorithms available, of which many are free. Such software has been referenced on the software before, but I believe there’s merit in having software that runs locally.
I’ve had some luck with Ultimate vocal remover, free software that allows stem separation to remove vocals, but depending on the model, also different instruments. It works best on computers with a somewhat capable gpu. The results are (to my ears) usually remarkably decent.
There’s quite some information out there already, so I won’t dive into detail too much. To start, just drop an mp3 and choose which tracks to extract.
I’ve had luck with the following settings:
Process method: Demucs
Choose Demucs model: v4 | htdemucs_ft
Choose stem(s): Drums
and then the box “no drums only” (you can deselect if you also want an mp3 with the drums separated).
That’s basically it. Step 2 is frustration because the song was harder than I thought!
A version that runs online is MVSEP, at MVSEP - Music & Voice Separation.
It uses some of the same models, but the free version has some limits.
edit: found there’s some reference to software on the boards already
I found karaokeversion.com to be a good source for this, but it does cost $3 a song.
i’m a beginner with finger drumming, less than a year here at QFG, and i’m having a blast. i’ve gone through the beginning and chunks of the intermediate. i finally got to the point where i was starting to play along with some songs, specifically, the “6 Famous Breakbeats” and “You Know I’m No Good”. i loved playing it, but i would do so along with the youtube video, so i’d hear the drum part in the background.
i found with karaokeversion, it had all instrumental/vocal parts and i could decide which tracks to include for a backing track, then save that version and use it for practice. for the very few songs i was playing with, this works out great.
yes, it’s $3 per song, and they might not have the song you’re looking for, but when i weigh that against learning how to create the backing track myself, i’m okay with that at this point in my journey.
i sound like i’m plugging the site. i’m not. i just found it useful for me. : )
Nice, I was just looking at Mixcraft after it was mentioned in another post. The Pro version has a stem separation tool and splits them in the DAW ready to go by the looks. Seems pretty handy. Most of the music I listen to though is via steaming so not sure how I’d go using any of these.
I assume paid services like Mixcraft use models with similar performance as those available for free. That’s the main reason why I opened the post: Free stem separation/karaoke models are available, but it takes a little investigation at times.
