Subject:help needed on noise reduction on long file

Posted by: saultopaul

Date:9/12/2013 4:37:35 PM

hello i have downloaded trial of SL . I have a very poor recording of a 70 min talk lots of noise echo etc . AM desperate to get it to a level that I can hear the words and transcribe it. have used izotope but not quite there. WHen i can afford it i hope to purchase the training video. I find the sl very hard to understand. Izotope and Audition much easier. . I can't even figure out how to apply NR to the whole file ! Maybe as its so slow on my imac i might have to chop the file into ten minute sections . Any advice would be massively welcomed and greatly appreciated Than you Paul
ps it’s a shame someone doesn’t offer a live tutorial on Skype. I would be happy to pay for it!

Subject:RE: help needed on noise reduction on long file

Reply by: Cielspacing

Date:9/13/2013 2:34:10 PM

There are several webinars available on SL:

-The introductory one:
http://www.sonycreativesoftware.com/spectralayersintrowebinar

-The 1st advanced:
http://www.sonycreativesoftware.com/audioproblemswebinar

-And the recent one about SL2 new features:
http://www.sonycreativesoftware.com/spectralcastingmoldingwebinar

In the case your internet connection is not fast enough to keep up with the video flow, you may download them, by clicking on the "Download the webinar" link, if nothing happens you may right-click on the same colored link and choose the "save link As..." option. It should work, just bear in mind that these files are heavy.

Subject:RE: help needed on noise reduction on long file

Reply by: Sneddy

Date:9/13/2013 8:11:36 PM

Paul, I just purchased Waves WNS. I have SL V1 and am clueless as to how to make it work. I have several posts on this forum expressing my discontent with SL so I won't bore you here with more. I'm generally a big fan of SCS (Vegas, Acid, DVD Arch, CD Arch). Even if SL gets to be half as good as these other programs it might end up being useful - we'll see.

If there are any SL users reading this and you have had an ounce of luck with SL, I'd love to hear about it (especially if you were able to apply it to a stereo music track raising/lowering an instruments volume without artifacts.

Paul, If you upload a segment of the 70 minute talk, I'd be willing to see if WNS helps.
Sneddy

Subject:RE: help needed on noise reduction on long file

Reply by: UKharrie

Date:9/20/2013 8:02:07 PM

Isn't the issue here that we are familiar with conventional Audio fix-software . . . which is pretty poor, be honest . . .Whereas SL is a new way to achieve what couldn't be done before?

For SL to be a great success, there need to be some staged-tutorials which allow users to progress towards "Super-User"

Perhaps an expert can tell us, - but I'd think OP's chances of removing ECHO will be small. My logic is that the frequency will be identical to the wanted sound.

However, as its "fingerprint" is already present in the file (starting at the beginning).... but it would be a very clever trick to make a rolling-removal.

I presume (can OP tell us) that the audio-track was recorded at some distance from the performance, so there were multiple paths for the recorded sound. Am I right in thinking the "fix" (next-time) is to place a sony SDHC recorder nearer to ONE speaker - the only snag I foresee is that electrets may not like being overloaded by high SPL.

Subject:RE: help needed on noise reduction on long file

Reply by: Layerry

Date:10/10/2013 1:35:13 AM

If you use izotope or audition for broadband noise reduction, then you can try SLP to separate the speech from the reverb. Mind you, this is a syllable by syllable process.

Set up a new layer for your speech. Set up a second layer to attenuate by setting it to reverse polarity and -3db volume slider.

Finding the proper window size setting is critical. A higher value like 8192 will show frequency information more clearly, while a lower setting will show time impulses more clearly. You want a window setting where the visual depiction of speech harmonics is slightly more blurred vertically, rather than horizontally.

Select the speech layer, and choose the harmonic extraction tool (5). Choose a short time length like 10ms. Set the bins to 1 to narrow the frequency range of your selection. Use however many harmonics you need. More is not necessarily better, but somewhere between 10 and 20 should work.

Next draw a time selection (T) around the area with which you would like to work.

Now look at your first syllable. Find the brightest partial and count up from the bottom. Set your harmonic to this number. Move your cursor over the bright partial. Press the TAB key to hear your selection. It should sound natural. If it doesn't, move to another partial and try it again. Finds a place on the partial where you get the least amount of reverb tail. When you find it, left click it to extract it to the layer.

Work through all of the syllables in your time selection this way.

All the mud is in the bottom. To find it, invert the speech layer (-). Now select your attenuation layer. Use the area extract tool (3), to brush the bottom below 600 Hz. Be careful not to remove any plosives like B's, P's, or D's.

If you now mute your speech layer, it will reappear at its normal volume, minus the muddy reverb tail.