I am running the newest version of audacity (installed this morning via .exe installer) on Windows 8.1. I have to transcribe a two hour audio for work. The gentlemen are in a car. Their voices are already extremely low and there is a consistent, recurring car noise. I need to know how to remove the car noise and isolate and bring up their voices ASAP. I have to get this done and I have tried everything. VERY FRUSTRATED. Any help would be greatly appreciated.

Link to audio sample: https://www.dropbox.com/s/ag6aieh1aabx733/Forum%20Sample.wav?dl=0

Please, please help.


I have to get this done

What is “this”? Do you need to make some improvement? (Maybe possible.)

Are you trying to make something intelligible that’s currently unintelligible? (Probably impossible.)

Are you trying to make a “studio quality” recording? (Impossible.)

and I have tried everything.

Again, what is “everything”? Have you tried [u]Noise Reduction[/u]? It’s worth a try, but if the noise is bad, “the cure can be worse than the disease.” Noise reduction works best when you have a constant low-level background noise… It works best when you don’t really need it.

You can also try high-pass filtering. A high-pass filter at around 200Hz will reduce/remove low-frequency noise while having little effect on the voices.

You can try Equalization to boost the higher frequencies (maybe above 3000 or 4000 Hz). That will boost the “T” and “S” sounds which can help with intelligibility. But of course, it will also boost high-frequency noise. And otherwise just play-around with the Equalizer to see if you can boost the voice frequencies and/or reduce the noise-frequencies. There’s a LOT of overlap, so this won’t work perfectly. When experimenting with the Equalizer, I recommend the Graphic EQ mode, instead of the Draw Curves mode.


Yeah… That’s why pros still record in soundproof studios, with good equipment, good microphone position, and good-professional talent… There’s only so much you can do with software… Even with good-professional software…

No. That’s the end.

Step One in Noise Reduction is to identify a segment of the work that’s noise only—pure noise without the voices so the tool knows what to “chew on.” There isn’t enough performance in your sample to tell when they stop and start talking.

There is no show. Time for a reshoot.


Digging out voices from a sea of noise is a specialist task requiring expertise and specialist equipment. If this is an important or legal matter, you could try seeking advice from an “audio forensics” professional.

Recording people without their consent is illegal in some countries, and considered morally dubious in most others, but if this recording was made with the permission of the people talking, then, as koz wrote, you will need to re-record it with much better microphone placement (for example, with headset mics).

As we are not able to offer practical advice about improving this recording, I will close this topic.