How to improve the sound quality of an mp3 speech file?
I have an old mp3 recording of a talk once given by me (recorded by an old-fashioned MP3-player). It has too low voice, too much noise, and too much crackling. Is there an application with which I can easily (i.e., without going too deep into physics) improve the sound quality?
Audacity 
Audacity has many good audio processing tools suitable to improve the sound of poor audio recordings.
- Noise reduction
- Amplifying without distortion of loud parts
- Removal of peaks
Noise reduction/removal
-
Load audio file into Audacity for processing
-
Select an area containing background noise only
-
Open Effect > Noise Removal...
-
Choose Get Noise Profile to use the previous selected noisy area for filtering.
-
Select all of your audio sample to apply the filter.
-
Choose Effect > Repeat Noise Removal.
Background noise was now effectively removed from our sample. -
Listen to the sample for artifacts in case we were overdoing denoising. If so undo your work and repeat steps 2. to 6. using different settings until we are happy.
-
Export the processed audio in any audio format or compression you wish.
Be gentle with your sound.
Audacity's denoising algorithm has the advantage to selectively filter the background noise but it will also filter these frequencies from our original. Therefore we need to be gentle with settings. Sometimes it may be better to leave some noise in the sample to avoid a too much distorted sound in the denoised audio file.
In the audio Effect section of Audacity we will find more useful filters for audio improvements:
Compressor
The compressor will increase low volumes and decrease the volume of louder parts in our audio. Again, this effect will be applied to a selection of our audio files.
Normalize
Normalizing audio will try to stretch the volume range over possible values. Unlike a compressor (above) it will preserve the dynamic of a recording.
Repair
In case of very short artifacts (clicks or the like) we can try to remove these wit the repair tool. This tool will try to interpolate the audio values from before and after the artifact to remove a tiny bit of our sample (less than 128 sample only).