"Restore frequencies" effect?

[Dalal]

Let's say I take an audio clip and completely wipe out all frequencies above 1000hz. Are there any plugins that will attempt to 'restore' the clip to its original state by intelligently guessing the high frequency content?

[testkid008]

That´s not exactly what it is doing, but have a look at http://www.zynaptiq.com/unfilter/... Watch the videos!

[BertKoor]

Usually what's not there cannot be put back.
But closest to what you describe is an Aural Exciter such as made by Aphex, also goes by the name of Sonic Maximizer by BBE. The main principle is applying harmonic distortion, but only add the highest audible octave of that. It's like MSG for music. Some producers love it, others can't stand it ...

[Dalal]

@testkid008, UNFILTER is pretty impressive, thanks for sharing. Looks like it doesn't quite work for my experiment, which is to strip away higher frequencies from a vocal and digitally reproduce them for a somewhat electronic but clean sound. UNFILTER doesn't sound too great. Either my experiment failed, or it's not the product I'm looking for.

@BertKoor, thanks for sharing the exciters. I always stayed away from exciters, but perhaps I should look at them again because in principle they do seem to be doing some of what I want. Perhaps the sound I expect out of this process is different from what is the reality.

[Shy]

Let's say I take an audio clip and completely wipe out all frequencies above 1000hz. Are there any plugins that will attempt to 'restore' the clip to its original state by intelligently guessing the high frequency content?

No, there is no plugin that does that. To intelligently build high frequencies from a source with only lower frequencies, you need something like a perceptual audio coding format which provides the basis for an SBR (spectral band replication) system which provides the basis for a high frequency regeneration (HFR) system based on a sine waves + noise model. The audio format itself (Musepack, Vorbis, AAC etc.) provides the values necessary for high quality "reconstruction", like separate fundamental frequency / harmonics and noise components, phase, amplitude, etc. The problem is that there is no implementation available publicly which is based on high quality HFR (only SBR like HE-AAC) and which enables "reconstruction" extreme enough to be significant for very low frequency sources (below 5 kHz), although it could be done, but I don't think anyone would bother.

Other methods like hybrid ADPCM-like coding, non-perceptual coding (lossless formats), plain processors which distort / generate harmonics ("exciter"/"maximizer"), etc., don't give you highly intelligible high frequency regeneration. And of course, a processor which only equalizes frequencies (that includes the Unfilter plugin) can't generate / synthesize anything. You also mentioned you want to cut frequencies at 1000Hz. That's likely too low for the source to be intelligible in the first place. Even telephone systems' frequency range extends up to about 3.4kHz and almost all audio going through telephones nowadays has up to at least 2.5kHz, and that's with speech-specific filters that improve intelligibility in the speech range. If you're aiming for a weird result, it may be good. But a "clean reconstruction" is simply not possible if your source only has a very limited frequency range.

Various "non-clean" methods which you may find surprisingly useful and nice are:
- filter bands through "envelope followers" through a noise generator, as found in various vocoders
- various codecs with different bit rates (can be used as "bit-crushers" with different flavors, see Chipcrusher)
- distortion such as that in CDP, especially the "Exaggerate" type, which exaggerates the contour of wave cycles as opposed to clipping waveforms / overloading. The result can be pretty amazing and much more useful than Exciters' or whatever.
- manually setting up high quality pitch shifting by different octaves and filtering each result appropriately separately, then combining all of them. You'd be surprised what can be done. But none of that is quick and simple.

[camsr]

It's like inductive VS deductive reasoning:
One assumes facts, the other knows.
You can add frequencies all day but they will not be what was originally there, unless you add what was originally there.