Help me achieve an audio effect.

February 11, 2006 11:56 PM

Audio Filter: I want to achieve an effect but don't know how. Is there a term for this / easy way to do it?

I have two clips, voice and wind. I want to combine them so that they sound like someone is speaking wind. I think I can achieve this by somehow mapping the volume of the wind clip to the waveform of the voice clip. Almost like a overlay matte in video. Is there an audio term for this? Am I heading down the right path? Suggestions?

posted by nathancaswell to Media & Arts (12 answers total)

I don't know a name for it, but the "Acoustic Mirror" function in Sony Sound Forge can achieve this, with one file as the "impulse" file, and one as the file being edited. Try searching for something like "sound morphing" and the like.
posted by fake at 12:31 AM on February 12, 2006

Simple volume (velocity, level) mapping won't give you what you're after. You want to check out either convolution or vocoding, both methods for "multiplying" one audio file by another so you get a mixture of the qualities of both files, not just the level of one applied to the frequency or spectrum of the other. The tools I’d check out to do this are all pricey and mostly Mac-based (Logic, Peak, SoundtrackPro, MetaSynth, Vokator, Reaktor, Kontakt), but I've no doubt free apps exist to experiment with them, too, for every platform. Start your search here.
posted by dpcoffin at 12:34 AM on February 12, 2006

Acoustic Mirror is a PC convolution app.
posted by dpcoffin at 12:36 AM on February 12, 2006

I've used the freeware app SoundHack for exactly this sort of thing before. Unfortunately it's Mac only, but it does allow for all sorts of cross-synthesis / FFT based convolution effects.
posted by Jon Mitchell at 1:10 AM on February 12, 2006

vocoding and convolution will work, but with a sound source such as wind, the results might sound pretty rough ... if you don't have to use that particular voice clip, you might try recording yourself whispering the phrases you want the wind to say and mixing the two files to the right proportion and then using a noise gate on the mix so that only the spoken parts are heard ... that might work also

whispering and wind are both sounds with a lot of pink noise in them, which is why vocoding wind might be tricky ... but mixing whispering and wind just might work ... it's an alternative to think about if you don't like the results you get from vocoding
posted by pyramid termite at 6:18 AM on February 12, 2006

What you really want to do is sound morphing, and unfortunately, that's hard. I know that Kyma can do it, but I am unaware of anything cheap or free that can.
posted by Eamon at 6:58 AM on February 12, 2006

Response by poster: Luckily, I have a Mac and a PC, so platform isn't an issue. Thanks for the pointers.
posted by nathancaswell at 8:11 AM on February 12, 2006

Probably the best thing to do is hand ride a filter. That way you can do it in the rhythm and cadence that you want.

See our mouths are nothing but filters, that is what enables us to make vowels - consonants - words. It's all the same source of nose coming from out throat, but it's the way we shape our mouths that makes it any kind of sound.

So it would go like this if I was to do it in Reason. You could use the same techniques in any other software studio/DAW, but I think reason is the easiest for folks to get their heads around.

1. Load up your wind sample in a sampler.

2. Attach assloads of filters to sampler in parallel.

3. Then draw the shapes of the sound you want on the tracks for the filters.


btw, there are filters in most of the machines in reason. don't just stick with the *filters.* If


On the more technical track you could use Max/MSP or PD and really start to have some fun!
posted by bigmusic at 8:13 AM on February 12, 2006

Like others have said, a vocoder might do what you want. Self link here - I wrote a simple vocoder program that runs on most platforms. Get it here.
posted by Emanuel at 9:03 AM on February 12, 2006

Another vote for using SoundHack's convolution feature.
posted by Dean King at 1:20 PM on February 12, 2006

I have to disagree with the vocoding/ convolutional line of thinking here. If you want any pink noise to sound link it is "talking" all you need to do is *shape it* with filters -- this is the easiest way to do any kind of vocal sounding synthesis. Vocoding will sound nasty and electronic and convolution isn't meant for doing that kind of treatment to sound.

This is simple filtered pink noise, and here is the source file for reason. This was done with a band pass and varying the freq cutoff and the resonance. You could make even more complicated filters to duplicate vowels by putting a couple bandpasses in series.

It might not work so well in reason because of the "stepping" -- this is not because the concept itself wouldn't work, it's because reason only has 128 points that you change on each variable (to match midi implementations), so it's not "smooth." You could probably do better thru PD, as I don't think it has that limitation.
posted by bigmusic at 2:00 PM on February 12, 2006

Antares kantos is an audio-controlled synthesizer that extracts pitch, volume, and formant data. I know you can apply the data to synth output, but I don't know if you can modify an another sound with it.
posted by abcde at 6:56 PM on February 12, 2006