PDA

View Full Version : Re-encode audio to filter out severe static?




iphoneZ
Apr 3, 2013, 02:37 AM
i am trying to watch a series of videos on youtube that were apparently recorded with really bad audio. (ironically, on the topic of information theory...)

severe static spikes throughout the video. if I download the videos any ideas on how I could reprocess them to cut out the static? seems it should be possible ... if I can do this on the command line, eg. using mencoder, that would be a big plus, as I'm not going to have time to do this manually.

here is an example of what I am talking about.

WARNING: very loud painful static kicks in about 3 seconds after the starting point of this video:

http://www.youtube.com/watch?v=imZRqcXehCI&t=96

if you can point me to another forum that might be able to answer this question that would be great also.



ChrisA
Apr 3, 2013, 05:20 AM
i am trying to watch a series of videos on youtube that were apparently recorded with really bad audio. (ironically, on the topic of information theory...) ....

Well information theory tells me you can't remove white noise.

BUt you can do two things. Use a DAW like Garage Band. Import the audio file then use volume automation to "ride" the volume control and mute the audio when you can

The other thing is to play with the EQ. Boost the vocal range and cut the highs and then maybe you can hear voices over the static

But in both cases you need to run the controls in real time. Automation helps

iphoneZ
Apr 3, 2013, 11:58 PM
to clarify i'm not hoping to recover the voice signal. I would be happy to just zero out the volume during those parts. (so that if i'm listening i don't get blasted by static.) but it's gotta be automated to be useful to me

cheekypaul
Apr 5, 2013, 02:55 AM
to clarify i'm not hoping to recover the voice signal. I would be happy to just zero out the volume during those parts. (so that if i'm listening i don't get blasted by static.) but it's gotta be automated to be useful to me

i'm not sure any automated set up would 100% successfully remove the noise. however, if you can see the waveform on a screen, the static will be visually very easy to spot, and therefore it wouldn't take long to select the bursts (and quickly see how many there are) and simply silence / lower the volume.
an automated set up presumably would need to 'listen' to the whole lecture, where by examining the waveform can be done faster than real time...

no use?