2

I have some clips from a Dropcam, and I'm trying to improve the audio to make some soft human speech discernible.

The speech is quite soft and fuzzy sounding, as the speakers are away from the camera (and it's a Dropcam as well, so I'm guessing the audio quality isn't superb, plus it's AAC encoded)

The audio is mostly human voices, no music - and the important thing is to make what is being said discernible - so I'm OK if it affects the timbre. It's a quiet environment, and there should be little background noise - the voices are just very soft/fuzzy.

I tried using mpv, and the -softvol flag (going up to around 400%) and the voices become quite artificial and hard to make out at times. And there's a hiss, not sure why (I thought there wasn't much in the background, but maybe I'm wrong).

Are there any tools or techniques that you would recommend to boost the volume, and improve the understandability of the voices and remain somewhat natural?

I've provided a AAC sample here (there are ones which are even software as well), in case that makes it clearer:

https://drive.google.com/file/d/0B70aFYh1VqvzaGNrcDFBSHFIUGc/view?usp=sharing

Improve this question
3
  • Not sure about this... The voices are incredibly hard to hear in that AAC file you've included. Amplifying it would only make the background noise and hiss even louder making it even harder to hear the voices which are completely indiscernible to begin with. – Daniel May 11 '16 at 23:50
  • A movie or TV studio uses post-production to dub over weak sounds. The boom mic acts as more of a queue for the timing of the sound. Without dubbing your limited to whatever DSP engine you have to extract the vocals from the background. – user18060 May 12 '16 at 21:58
  • Unless the camera is in their faces, I suggest getting a boom and quality 'directional' mic and treat the video and audio as separate issues to deal with in post. The boom mic is usually over the 'actors' heads high enough to be out of the camera's view. Your 'base' sound quality should improve a lot. – user18060 May 17 '16 at 22:41
1

Unfortunately boosting the signal you have is unlikely to help - where you don't have the information in the signal you can't really create it.

As you have already seen, while DSP's can do a lot to increase the distinction between noise floor and signal, and between unwanted frequencies and speech you are ultimately limited by the quality of the recording.

Similar to Sparky256's comment, I'd suggest your only real option is to get enough clarity that you can understand the lines, and then overdub them in post.

Improve this answer

Your Answer

By clicking “Post Your Answer”, you agree to our terms of service, privacy policy and cookie policy

Not the answer you're looking for? Browse other questions tagged or ask your own question.