Hi.
I looking for a help of passionate hobbyist of audio cleaning/restoration. I need a help in cleaning and extracting voice conversation from recording with has a background noise (some water) and TV set in background.
The conversation is not in english language (it is in Polish).
I uploaded 4 minute fragment if you want to help me, download link 10 MB => https://drive.google.com/file/d/0By-3Ux ... NyTEE/view
Here goes the full recording 50 minutes 222 MB=> https://drive.google.com/file/d/0By-3Ux ... VkMzQ/view
The other option is to help me with Izotope RX4 or Adobe Audition 2014 to do this myself. I know how to use any Windows application and I'm IT guy, but I have no experience (except simple audio edition or WAVE/FLAC/MP3 conversion) in audio de-noising, de-clicking, restoration etc.
I hope there is someone who has experience in increasing vocal/conversation from audio files and will help me.
Thanks in advance
Jacek
Looking for hobbist willing to help in cleaning audio, exchanging/increasing conversation
-
- KVRAF
- 16866 posts since 8 Mar, 2005 from Utrecht, Holland
Cleaning up static noise (hiss) from recordings is difficult enough, I wish you all the luck with that water stream and a TV in the background.
There are several strategies I know to clean up recordings:
1) A noise gate can make the signal silent (or just lower the gain with some dB's) when the audio level drops below a certain threshold level. Works quite good if the noise is faint, but I find in practice this destroys the sound more than it improves it. You hear the noise gate opening & closing all the time.
2) Narrow-band filtering can be used to get rid of for example 50 Hz hum. You'd also need to add filters to get rid of 100, 150 & 200 Hz harmonics as well which are also present. For example tape hiss above 12kHz can be filtered out also in a simular way with a shelf filter, but then you also lose some of the audio in that part of the spectrum as well.
3) With a short piece of recording containing just the noise (hum & hiss) you have a spectrum fingerprint of the noise. Eg Adobe Audition can use such a selected fingerprint to recognise noise and filter that out using FFT processing. Works quite well if you set this noise remover algorithm not too aggressive and wash out the noise in several passes.
The noise you have does not fall in any of these categories. Maybe there are tools that can do it, but I don't know about them.
Part of the problem is that our own brain is quite good at analysing sounds & identifying the separate components you hear. But it took 2 billion years of evolution to get there. You can't expect a computer to do the same. Alas we're not there yet. Slowly we're getting at a level where it can recognise the pitch of a voice, but in a full mix it becomes difficult. Your problem is a bit simular to vocal isolation, and last time I looked we weren't there yet by far.
There are several strategies I know to clean up recordings:
1) A noise gate can make the signal silent (or just lower the gain with some dB's) when the audio level drops below a certain threshold level. Works quite good if the noise is faint, but I find in practice this destroys the sound more than it improves it. You hear the noise gate opening & closing all the time.
2) Narrow-band filtering can be used to get rid of for example 50 Hz hum. You'd also need to add filters to get rid of 100, 150 & 200 Hz harmonics as well which are also present. For example tape hiss above 12kHz can be filtered out also in a simular way with a shelf filter, but then you also lose some of the audio in that part of the spectrum as well.
3) With a short piece of recording containing just the noise (hum & hiss) you have a spectrum fingerprint of the noise. Eg Adobe Audition can use such a selected fingerprint to recognise noise and filter that out using FFT processing. Works quite well if you set this noise remover algorithm not too aggressive and wash out the noise in several passes.
The noise you have does not fall in any of these categories. Maybe there are tools that can do it, but I don't know about them.
Part of the problem is that our own brain is quite good at analysing sounds & identifying the separate components you hear. But it took 2 billion years of evolution to get there. You can't expect a computer to do the same. Alas we're not there yet. Slowly we're getting at a level where it can recognise the pitch of a voice, but in a full mix it becomes difficult. Your problem is a bit simular to vocal isolation, and last time I looked we weren't there yet by far.
We are the KVR collective. Resistance is futile. You will be assimilated. 
My MusicCalc is served over https!!
My MusicCalc is served over https!!
-
- KVRAF
- 2989 posts since 5 Nov, 2014
Recently I had similar task and got hold on it pretty fast I think, I spent most of my time actually comparing/demoing different solutions, than actually working, by the middle of that pile of job I got hang on it and pretty much without much trial&error had workflow/routine/chain that kinda worked, I felt it was decent enough, other party used better adjectives, so both sides were happy.
Here's my thread where I started the quest and you can find some of my initial findings
http://www.kvraudio.com/forum/viewtopic.php?p=6049476
Acon Digital is the way to go, if you have IZotope and Adobe solutions already, that's great too, but if not, Acon Digital just provides great results, their Acoustica Premium is the way to go or if you are more comfortable with some other editor, than their Restoration Suite
http://acondigital.com/
Here's my chain:
-First I do DeNoise/Hum/Click/Clip-ing, whatever is needed, pretty straight forward stuff and you will get hang on it fast, I think after this video you are good to go
-Surgical EQ mostly HP/LP-ing unnecessary ranges like anything bellow 100hz or above 14-16khz
-Multi band compression, aim here is to push the midrange up, I accidentally stumbled upon preset and felt it was bringing something to the table, so he became part of the chain, here's the picture
-Next thing was boosting and cutting, from everything I tried for particular task, graphic EQ was the bomb, these settings worked for me, it may be too much for you, so experiment and use your ears, but these work kinda good on material I had for numerous reasons, so really YMMV
-Next was De Ess-ing, not needed that much for some reason, but your YMMV
-Next was limiting, I used two limiters in a row, not heavy lifting anything or giving them much trouble, so results were good, but you can work with one or none, really it's up to you
Worth pointing out, there was fair amount of peaks, just on few places, bring them down manually, take the time, that's essential for getting most out of limiter and just having louder outcome without distorting.
Just get in everything that needs more work, I feel DeNoise-ing is essential here and better results from it, less work you got, but that's my findings and for that particular reason I can't recommend enough Acon DeNoiser, it's the thing for the job, really delivering great results.
I see you will going to use DeClip-er too based on that short recording you left.
If you have any questions, shoot, whatever the term or anything, I will be checking out this thread or if you feel more comfortable shoot me PM, I'm noob as far as this goes, just did few gigabytes of work on old cassette converted field interview recordings, but I'm willing to help out as much as I can and know, not much to offer, but hey
One other thing, I was working with these, they are good for the job, pretty cheap and delivering, if you by any chance need something suitable for the task, these are the ones
http://en-us.sennheiser.com/over-ear-headphones-hd-201
Here's my thread where I started the quest and you can find some of my initial findings
http://www.kvraudio.com/forum/viewtopic.php?p=6049476
Acon Digital is the way to go, if you have IZotope and Adobe solutions already, that's great too, but if not, Acon Digital just provides great results, their Acoustica Premium is the way to go or if you are more comfortable with some other editor, than their Restoration Suite
http://acondigital.com/
Here's my chain:
-First I do DeNoise/Hum/Click/Clip-ing, whatever is needed, pretty straight forward stuff and you will get hang on it fast, I think after this video you are good to go
-Surgical EQ mostly HP/LP-ing unnecessary ranges like anything bellow 100hz or above 14-16khz
-Multi band compression, aim here is to push the midrange up, I accidentally stumbled upon preset and felt it was bringing something to the table, so he became part of the chain, here's the picture
-Next thing was boosting and cutting, from everything I tried for particular task, graphic EQ was the bomb, these settings worked for me, it may be too much for you, so experiment and use your ears, but these work kinda good on material I had for numerous reasons, so really YMMV
-Next was De Ess-ing, not needed that much for some reason, but your YMMV
-Next was limiting, I used two limiters in a row, not heavy lifting anything or giving them much trouble, so results were good, but you can work with one or none, really it's up to you
Worth pointing out, there was fair amount of peaks, just on few places, bring them down manually, take the time, that's essential for getting most out of limiter and just having louder outcome without distorting.
Just get in everything that needs more work, I feel DeNoise-ing is essential here and better results from it, less work you got, but that's my findings and for that particular reason I can't recommend enough Acon DeNoiser, it's the thing for the job, really delivering great results.
I see you will going to use DeClip-er too based on that short recording you left.
If you have any questions, shoot, whatever the term or anything, I will be checking out this thread or if you feel more comfortable shoot me PM, I'm noob as far as this goes, just did few gigabytes of work on old cassette converted field interview recordings, but I'm willing to help out as much as I can and know, not much to offer, but hey
One other thing, I was working with these, they are good for the job, pretty cheap and delivering, if you by any chance need something suitable for the task, these are the ones
http://en-us.sennheiser.com/over-ear-headphones-hd-201
-
- KVRAF
- 2989 posts since 5 Nov, 2014
You are welcome, if you have any further questions/doubts/troubles, I'm here for you 
-
- KVRian
- 1100 posts since 9 Jan, 2015 from NY, NY
Can't listen to the clip as I'm at work, but one option is to use the noise removal function in Audacity (http://web.audacityteam.org/)
If you can find a section with just the noise in the audio you can sample it and Audacity with remove that noise from the entire file. It's not perfect, but in some cases it can do a good job at removing most of the noise.
Here's the manual entry for it: http://manual.audacityteam.org/o/man/no ... ction.html
If you can find a section with just the noise in the audio you can sample it and Audacity with remove that noise from the entire file. It's not perfect, but in some cases it can do a good job at removing most of the noise.
Here's the manual entry for it: http://manual.audacityteam.org/o/man/no ... ction.html
Sweet child in time...
-
- KVRAF
- 2989 posts since 5 Nov, 2014
Audacity is pretty bad at noise removal IMHO, too much artifacts, for free little better solutions as far as noise removal goes are WavePad and Nero WaveEditor, but take that with grain of salt, these are my personal findings/taste that are based on particular material, YMMV.
http://www.nch.com.au/wavepad/
http://www.nero.com/enu/downloads/
Also, neat basic audio editor, maybe just someone's cup of tea
http://www.ocenaudio.com.br/
http://www.nch.com.au/wavepad/
http://www.nero.com/enu/downloads/
Also, neat basic audio editor, maybe just someone's cup of tea
http://www.ocenaudio.com.br/
-
- KVRian
- 1100 posts since 9 Jan, 2015 from NY, NY
Interesting... I have used Audacity successfully to remove buzz and hum, but not really experimented with other noise too much.
I'll try these and see if they work better for anything I come up against.
*Edit*
Is this one free or just a demo?
http://www.nch.com.au/wavepad/
I'll try these and see if they work better for anything I come up against.
*Edit*
Is this one free or just a demo?
http://www.nch.com.au/wavepad/
Last edited by Deep Purple on Fri Apr 24, 2015 3:17 pm, edited 1 time in total.
Sweet child in time...
-
- KVRAF
- 2989 posts since 5 Nov, 2014
It's a plugin, if you don't have one within your audio editor, I would recommend checking out ReaXcomp or C3 Multi Band Compressorjcsjacekj wrote:Passing Bye
I stuck on:
- Multiband compression - it is a plugin or standalone product?
http://www.kvraudio.com/product/reaxcomp_by_cockos
http://www.kvraudio.com/product/c3_mult ... er/details
It's free for non commercial useDeep Purple wrote:Interesting... I have used Audacity successfully to remove buzz and hum, but not really experimented with other noise too much.
I'll try these and see if they work better for anything I come up against.
*Edit*
Is this one free or just a demo?
http://www.nch.com.au/wavepad/
For me, Acon Digital is just top, at least De Noise-er, IMHO better than popular and pricier iZotope, Waves, Sonnox and etc solutions
-
- KVRer
- Topic Starter
- 8 posts since 24 Apr, 2015
Passing bye.
I got the multiband compressor. You wrote "aim here is to push the midrange up".
On you screent the midrange (orange & green bands) have - gain. So for mi ranges what gain i should choose (+ volume up or - volume down)?
I got the multiband compressor. You wrote "aim here is to push the midrange up".
On you screent the midrange (orange & green bands) have - gain. So for mi ranges what gain i should choose (+ volume up or - volume down)?
-
- KVRAF
- 2989 posts since 5 Nov, 2014
That's gain reduction, compensating for perceived volume loss caused by compression/pulling down stuff beyond threshold by set ratio amount in nutshell, if you want to learn more about it, here's nice onejcsjacekj wrote:Passing bye.
I got the multiband compressor. You wrote "aim here is to push the midrange up".
On you screent the midrange (orange & green bands) have - gain. So for mi ranges what gain i should choose (+ volume up or - volume down)?
Anyways, I don't think that compression is that important here, so don't put much importance on it or that whole chain as a must, rather just see to really do great job with cleaning/restoring and with EQ if you feel material needs it, but overall, keep it simple and effective, that chain may worked for me, but it may just be overkill for you or unnecessary
-
- KVRAF
- 2989 posts since 5 Nov, 2014
Here you go, I made you preset for it, just import itjcsjacekj wrote:I got this one, how to setup it?
What values for each band?
https://dl.dropboxusercontent.com/u/904 ... 0Vocal.zip
-
- KVRAF
- 2989 posts since 5 Nov, 2014
Also gonna just add that this tool is great for the job, was little heavy on my old crappy machine so I didn't used it, but that's really the tool for this kind of stuff and replaces graphic EQ and multi band comp in that chain
https://vladgsound.wordpress.com/plugins/nova67p/
https://vladgsound.wordpress.com/plugins/nova67p/