I got this vocals off a YouTube video and tried separating the instrumentals from it.
The only problem is, I can’t seem to find a solution as I’m a complete beginner and I also did some googling myself but to no luck.
I hope someone could help me and give me advices so I can do better in the future!
Here’s the audio I’m trying to fix together with the original song vocal I’m trying to replace it with. Vocal1 - the vocals I’m trying to fix - Vocal2 - the vocals separated from the Instrumental Instrumental - the instrumental
Which one is the mixed song? We can’t take processing out of a sound file, so we need what you started with not what you ended up with.
This seems to be a very highly processed song and those don’t make very good candidates for vocal separation. Some times people get lucky with one song and post the results on YouTube and then nobody else can get it to work on any other song. That’s normal.
You can try the classic Split Stereo, Invert one channel, and Add approach. Or you can use Effect > Vocal Reduction and Isolation which is basically the same thing with some added features. If neither of those works for you and that song may not be a good candidate for processing.
Most of the time it doesn’t work very well. It really doesn’t work very well if you start out with an MP3. MP3 compression messes with musical tonal values to get those small files and that may be what you’re experiencing in this song.
The Vocal2 and Instrumentals were originally from a single audio track and then separated using a 2-Stem Model separator website online. Same goes with the Vocal1 it’s taken from a karaoke cover of someone. I was mainly trying to make that karaoke cover better by using the original instrumentals together with the female vocals.
For my unprofessional ears, the instrumentals sound decent at least. I only really wanted to make the Vocal1 sound better somehow. Any additional tips? Thanks!
There are (free) plugins which increase the contrast of the spectrogram, like LEVELS & KONTRAST.
they can remove digital-artefacts and other unwanted noise.
Then add a generous measure of stereo-reverb which covers a multitude of sins …
Thanks for the suggestion! I tried playing with the sliders of the Levels and Kontrast but can’t seem to achieve the output you did. If only there was a decent tutorial. The description on the download site is kinda lacking for me.
Increasing the contrast is effectively noise-reduction, but you don’t need a noise-profile.
The most recent version, (2.0.3. , “Spectral Gate_Win32.dll”), works in Audacity on Windows,
(there is also a mac version).
Andrew Reeman’s “Spectral Gate” has an advantage over KONTRAST in that you can adjust the FFT size.
Larger FFT sizes produce a better result when processing singing with reverb.