Amplifying very very low audio

Any help would be appreciated…

I have a ZOOM H4, which has two input options - mic or direct input. For my last recording (total of 15 talks), I accidentally recorded using the mic, and not the inputs. The recorder was in a room next to the speakers, so it picked up the talks… just very poorly.

I don’t think I’ll be able to have it re-recorded. I am not an audio professional (otherwise, this wouldn’t have happened).

I’ve tried a few things, but nothing so far that I’m pleased with. I have both Audacity and Soundbooth.

  1. I tried straight amplification. 40 dB and it is audible, but the hiss in the background make it impossible to listen to comfortably. The words can be made out.

  2. Normalizing, then removing noise (24 / 230 / .05) creates a slightly more listenable audio track, but nothing I think someone would want to listen to.

  3. In Soundbooth, I clicked the ‘Louder’ button at the bottom, applied a 6.5 dB increase, then removed noise 100% / 20 dB. This has the best result so far… but again, it isn’t easy to listen to.

Either I have an audio file where I can hear the speaker, but there is awful noise… OR the speaker comes through somewhat OK, but his voice sounds like he is a robot (the audio sounds like very high-compression on a compressed audio file).

So… is this sort of audio recovery impossible? I know there is no way to remove all audio defects, but I’m hoping I can salvage the audio.



Amplification and Noise Removal are the correct tools for the job.
There is a limit to how much improvement can be made, but if you could upload a SHORT sample (just a few seconds) that includes both “noise only” and “voices”, then we may be able to give some suggestions of settings to improve the result. (don’t expect miracles :wink: ) Upload the sample as a WAV or FLAC file (not MP3).

My suggestion would be to remove all the frequencies above 4KHz to reduce the hiss, then amplify.
Audacity’s equalizer, in “effect” menu, will enable you to do this. The result will be phone-like but hopefully the speech will be more comprehensible.