Jump to content


Photo

How do I edit a video to boost an audio channel - human speech

Video/audio processing questi

  • Please log in to reply
10 replies to this topic

#1 g__day

g__day

    Champion

  • Hero
  • 7,902 posts

Posted 25 November 2014 - 07:28 PM

Hey all,

 

I have a Video file ~ 500MB = 8 minutes duration in .MTS format.  The sound has recorded two instructors feedback as they watched my last grading, as well as all the noise from both our gradings and a badminton championship that was really noisy.

 

Can folk recommend the ideal software and processing to lift and boost human voice and diminishing all other (dynamic and variable sounds)?

 

I have tried a trial of Adobe 2014 Audition (as I already have Adobe CS4 Premiere and CS4 Sound booth - never used either before).

 

Audition has advanced functions to remove voices - but not the exact opposite - keep voices and diminish everything else.  It does have an improve voices - but it only marginally helps.  I can hear my Instructors say Matt is doing really well there - he needs to - then its all to faint amongst the background cacophony of noise. I am going through online tutorials at present to see how you do basic sound scrubbing effects.  A google search reveals over 25 products to try and improve audio! http://www.hongkiat....-audio-editors/

 

 

If anyone has specialised skills in this sort of sound scrubbing or editing - or knows of a service that does this - love to hear your advice!

 

Many thanks,

 

   Matthew


Talent + Integrity = Atomic!

#2 Rybags

Rybags

    Immortal

  • Super Hero
  • 35,848 posts

Posted 25 November 2014 - 09:24 PM

Possibly some sort of equaliser function would work, attenuate everything outside the normal speech frequencies.

But then if a good part of unwanted background is also speech, there'll probably be trouble.

 

It might be as simple as exporting the audio and running through Audacity.  Plenty of video packages are fairly minimal on audio processing and concentrate mostly on visuals and editing.



#3 komuso

komuso

    Overlord

  • Atomican
  • 2,316 posts
  • Location:interstitial

Posted 25 November 2014 - 11:49 PM

Hey all,

 

I have a Video file ~ 500MB = 8 minutes duration in .MTS format.  The sound has recorded two instructors feedback as they watched my last grading, as well as all the noise from both our gradings and a badminton championship that was really noisy.

 

Can folk recommend the ideal software and processing to lift and boost human voice and diminishing all other (dynamic and variable sounds)?

 

I have tried a trial of Adobe 2014 Audition (as I already have Adobe CS4 Premiere and CS4 Sound booth - never used either before).

 

Audition has advanced functions to remove voices - but not the exact opposite - keep voices and diminish everything else.  It does have an improve voices - but it only marginally helps.  I can hear my Instructors say Matt is doing really well there - he needs to - then its all to faint amongst the background cacophony of noise. I am going through online tutorials at present to see how you do basic sound scrubbing effects.  A google search reveals over 25 products to try and improve audio! http://www.hongkiat....-audio-editors/

 

 

If anyone has specialised skills in this sort of sound scrubbing or editing - or knows of a service that does this - love to hear your advice!

 

Many thanks,

 

   Matthew

 

Can you send a few minutes of audio through to me? I will see what I can do for you. If you don't want to post it up here send me a PM
 

It's probably a tough ask by the way, not likely to be the kind of thing your average plugin will take care of.


Edited by komuso, 25 November 2014 - 11:50 PM.


#4 g__day

g__day

    Champion

  • Hero
  • 7,902 posts

Posted 26 November 2014 - 10:42 AM

Extracting the Audio only reduces it to a 105MB .wav file - how much of the action do you want - size and/or time wise?

 

Must say - after watching so much CSI where a techie pushes a slider then all the background noise on a sound recording becomes crystal clear - isn't probably reality.  My tape is has random event noise (quite loud) for the first minute - then the interesting background conversation goes on intermittently for the next 2.5 minutes - with only half of this at best being barely legible.

 

Very happy to send this on as an email or a USB key if you think you can help!


Edited by g__day, 26 November 2014 - 10:43 AM.

Talent + Integrity = Atomic!

#5 komuso

komuso

    Overlord

  • Atomican
  • 2,316 posts
  • Location:interstitial

Posted 26 November 2014 - 01:35 PM

All I would need is enough of the audio to get a clear idea of the sounds that need treatment. Wouldn't hurt to send the whole thing through I guess. 

 

Probably wouldn't make for good CSI watching if you see the tech click/drag/select/ok/adjust/undo/repeat/etc :P



#6 Rybags

Rybags

    Immortal

  • Super Hero
  • 35,848 posts

Posted 26 November 2014 - 04:38 PM

You could compress the audio for mailing - a few minutes at high bitrate MP3 or AAC wouldn't use much space.  Then just convert back to WAV for processing.



#7 g__day

g__day

    Champion

  • Hero
  • 7,902 posts

Posted 26 November 2014 - 10:32 PM

Well I grabbed about 90 secs where most of the interesting talking was and ran this thru Audacity's high and low pass filters and managed to get maybe 60% of the dialogue - which is really helpful. The file as stands is now a more manageable 11.5MB - so I may explore Dropbox to hold it in case any interested can work some magic on it - many thanks all, off to read how Dropbox works!


Talent + Integrity = Atomic!

#8 komuso

komuso

    Overlord

  • Atomican
  • 2,316 posts
  • Location:interstitial

Posted 27 November 2014 - 09:40 PM

The starting point would be to use noise reduction software to analyse a section of what you want removed, then apply. After that, if the vocals still need enhancing, you can use multiband compression, focusing on formants, to enhance the intelligibility of speech. Formants are generally vowel related sounds, points in the frequency range that both the ear and the voice box have evolved to be more sensitive to for easier communication. Formants occupy very specific ranges and are thus easy to isolate, and you can increase intelligibility in a way that is aesthetically pleasing without being overly destructive, unlike the results that equalising might produce. What exactly needs to be done, and how much, really depends on the material though.


Edited by komuso, 27 November 2014 - 09:44 PM.


#9 Master_Scythe

Master_Scythe

    Titan

  • Hero
  • 20,571 posts
  • Location:QLD

Posted 29 November 2014 - 11:27 AM

is there an audacity plugin for formants?


Wherever you go in life, watch out for Scythe, the tackling IT support guy.

"I don't care what race you are, not one f*cking bit, if you want to be seen as a good people, you go in there and you f*ck up the people who (unofficially) represent you in a negative light!"


#10 komuso

komuso

    Overlord

  • Atomican
  • 2,316 posts
  • Location:interstitial

Posted 29 November 2014 - 12:31 PM

is there an audacity plugin for formants?

 

None that I know of. Celemony Melodyne can run as a plugin and can be used to manipulate formants, but its for retuning, and won't help in this instance.



#11 @~thehung

@~thehung

    Guru

  • Hero
  • 8,790 posts

Posted 07 December 2014 - 05:10 PM

the ideal software: iZotope RX  - unless things have changed in the last few years (when i havent been paying close attention), its still an industry standard which has been blowing peoples minds since it arrived on the scene.  it can do things that by rights should be 'impossible'.

 

the key to how the most surgical of noise reduction tools work, is the ability to transform the waveform data to spectral information that can be viewed and manipulated as a graph of fourier frequency data vs time.   it is possible then, for example, to see the dull drone of a distant passing train as a smudge extending across the bottom of the frame.  you would also see any audible harmonics present as corresponding parallel lines further above this smudge.  you can then simply draw a box around the unwanted audio, and the software will effectively overwrite the area of the box with what was 'normal' before/after/around it, much in the same way you would work with a clone brush if visually editing the graph itself.  conversely, you can define what you want to keep, and work from there.  the AI is crazy-good at this.

 

as far as working with EQ, its generally useful to severely attenuate everything below around 400Hz, boost in the 2k-5k range, and then attenuate everything above 5k.  this will help to remove muddiness and hiss, but of course, not so much with random noises that tend to overlap vocal frequencies. 


no pung intended




0 user(s) are reading this topic

0 members, 0 guests, 0 anonymous users