How to clean up a baby's first words video
You caught it on a phone. The baby said something you want to keep forever, and there's a TV or a fan running underneath.
The setup
Remove
"ambient noise" ยท preset
Off
Step by step
- 1 Drop your video or audio file into SoundScrub.
- 2 Set the mode to Remove.
- 3 Pick "ambient noise" from the sound list.
- 4 Run it. You get the file back with that sound removed and the rest untouched.
Why this works
Remove mode with the "ambient noise" preset takes out the fan, the hum, and the room, and leaves everything else alone. Remove only takes away the sound you name, so it can't drop the baby by mistake.
Why the usual fixes fall short
A noise reducer takes a swing at the fan and leaves the TV. Muting the noisy section clips the word you were trying to save.
Another way
If the TV is the main problem and it's mostly a soundtrack, Remove with "background music" targets that instead. Talking on the TV counts as speech and will stay.
Honest heads-up
Don't reach for Keep only "people speaking" here, even though it sounds right. It keeps what the model recognises as speech, and babbling or a half-formed first word may not count, so you risk deleting the exact thing you wanted. Remove is the safe direction on footage you can't get back. Also, if the words were faint because you were across the room, clearing the noise won't make them clearer.
Common questions
How do I remove background noise from a baby's first words video?
In SoundScrub, set the mode to Remove and pick the "ambient noise" preset. That clears the fan and room hum and leaves every other sound untouched, including the baby.
Should I use Keep only to isolate the voices?
Not on this. Keep only holds what the model recognises as speech, and babbling or a half-formed word might not register, which would drop the moment entirely. Removing the noise is the safer direction when the footage can't be repeated.
Can it keep just the baby and drop me talking?
No. SoundScrub separates speech from non-speech, not one person from another, so it can't pick one voice out of several.
Other guides
macOS, Windows, and Linux. No subscription.