Stem separation, noise reduction and EQ: choose the right audio tool

Stem separation, noise reduction and EQ: choose the right audio tool

Stem separation estimates sound sources such as vocals or drums. EQ changes frequency ranges, compression changes level dynamics, and a noise gate reduces passages below a threshold. These processes solve different problems. Choosing by the unwanted sound alone can lead to an edit that also removes the performance you want to keep.

Try this workflow

1. For singing over a song, try the instrumental from Vocal Remover. For a drumless or bassless backing, use the four-stem mixer.
2. For excessive bass or a harsh top end, test a small EQ adjustment. Compare at a similar loudness and keep enough headroom.
3. For quiet room noise between phrases, test the noise gate with a low threshold. Listen carefully to soft words and fading notes before increasing it.

What to check before you finish

SurStudio does not have dedicated AI wind, crowd, breath, click or reverb-removal models. Its gate cannot distinguish speech from noise at the same level. Likewise, the basic mastering chain is not an AI repair service. Avoid repeatedly stacking processors without checking what each step improves.

A practical example

If a home recording has room hiss during pauses and muddy low mids during speech, use two small adjustments for the two issues: a cautious gate for pauses and an EQ cut for tone. If music is masking the speech continuously, those changes may be insufficient; an original voice-only recording is a better source when available.

More questions

Can a bass booster isolate the bass instrument?

No. It increases low-frequency content from all sources. Use the Bass stem to hear an estimated bass part, and check for other low sounds in that output.

Is reverb removal the same as adding less reverb?

No. Reducing an effect control before rendering is different from removing ambience already mixed into a recording. SurStudio adds creative ambience but does not include a dedicated dereverberation model.