Is there a way to "smear" audio in the time domain?

[David Kissam]

Imagine you are looking at a spectrogram. Then increase the FFT size. The frequency resolution gets better and the time resolution gets worse. This results in the display of every audio "event" getting smeared, or elongated, left-to-right (in time).

Is there a way to actually process audio in this way, such that everything gets smeared in time? I want crisp, fast changes to be slowed down, so that they become "smooth".

As I understand it, linear phase filters do this somewhat. However, I don't want to apply any EQ, I just want the time-smearing.

And if it's not clear from this post, I don't really know what I'm talking about, or whether this would even accomplish what I want it to. But I'd like to try it so I can find out.

[mjolnir]

DtBlkFX (VST effect) can do this I think. There might even be a preset.
However, in REAPER, A regular time expansion of a WAV clip can do it too.
Just drag the right edge of the WAV clip while holding the ALT key.
Also, in REAPER stretch markers could be used, I guess.

[David Kissam]

DtBlkFX (VST effect) can do this I think. There might even be a preset.

Thanks, I will try this.

However, in REAPER, A regular time expansion of a WAV clip can do it too.
Just drag the right edge of the WAV clip while holding the ALT key.
Also, in REAPER stretch markers could be used, I guess.

I don't actually want to slow down the audio overall. Imagine a recording of a metronome at 60 BPM. Each click of the metronome is a short, sharp event. I would like to (slightly) soften each click by lengthening it in both directions - before and after, so it kind of fades in and fades out. But the tempo would still be 60 BPM.

That's not literally what I'm doing but it's an illustrative example.

[zerocrossing]

DtBlkFX (VST effect) can do this I think. There might even be a preset.

Thanks, I will try this.

However, in REAPER, A regular time expansion of a WAV clip can do it too.
Just drag the right edge of the WAV clip while holding the ALT key.
Also, in REAPER stretch markers could be used, I guess.

I don't actually want to slow down the audio overall. Imagine a recording of a metronome at 60 BPM. Each click of the metronome is a short, sharp event. I would like to (slightly) soften each click by lengthening it in both directions - before and after, so it kind of fades in and fades out. But the tempo would still be 60 BPM.

That's not literally what I'm doing but it's an illustrative example.

Maybe this?

https://gs-dsp.com/products/magicblur/

[David Kissam]

Ah, I think I know a more precise way to say what I want.

I want to divide a signal into small frequency bands (something like what iZotope RX does), and get a moving average of the level of each band. Then, by adjusting the time interval of the average, I can change the amount of smoothing applied. Am I on to something or am I way off?

[J.Ruegg]

I want to divide a signal into small frequency bands (something like what iZotope RX does), and get a moving average of the level of each band. Then, by adjusting the time interval of the average, I can change the amount of smoothing applied. Am I on to something or am I way off?

Would vocoding the signal with itself accomplish that? You could adjust the attack and release times to make it behave slower to the input signal.
The downside is that I don't know a vocoder that has an absurd amount of bands.

I wonder if a spectral compressor with long attack times would be better.

[Ah_Dziz]

GRM tools evolve seems like the closest thing. Does pretty much this by interpolation between fft frames that are separated by relatively long periods of time. It can go to extremes where it's like the sound is frozen but always slowly changing but can also act quickly or be triggered by the input signal.
Other "spectral blurring" effects can be had by randomizing bin phase in a large fft. A spectral thresholding effect can also do this stuff. Adjusting the attack and release settings will exaggerate the effect further. Soundhack spectral shapers is a good one for this as is MSpectralDynamics. You could also use all pass filters or a resonator bank. Lots of ways to do a version of what you want.

[slartibartfast]

My suggestion would be to apply a reverb to the original track (play with the reverb settings to give the effect you want). Then reverse the original track, apply the reverb to that track with the same settings as for the first track, freeze the effected track and reverse this back again. Then overlay the two tracks
- the one with the reverb applied and
- the reversed, reverb-ed, and re-reversed track.

A bit of faffing involved, but see if this gives the effect you are after.

[Jorchime]

I wonder if you are really just looking for a low-pass/hi-cut.
If you filter out higher harmonies, you are filtering those frequencies that make up sharp edges in the time domain -> you get a smoothing effect.

I'm sure there is a nice impulse response you could use through convolution.
The dspguide has a couple of points with regards to average filters: https://www.dspguide.com/CH15.PDF
Maybe this provides some more hints on which terms to use / what to look for.

[whyterabbyt]

Google 'audio spectral blur'

[swilow11]

Unfiltered Audio Specops could maybe do this.

[zerocrossing]

Maybe something like Photosounder?

https://photosounder.com/

One issue that hasn't come up is that you mentioned blur, which works on a photograph because a photograph doesn't have any temporal component. So you can't put a blur on a sound in the same way, not in real time, because you're dealing with the arrow of time, which as far as we can tell, is a one way sign. So, you'd need a huge buffer, or to preapply the effect. Reverb is essentially audio blur, but it only blurs in one vector.

[David Kissam]

Thanks everyone for the replies. You've given me a lot of good ideas.

[osiris]

Maybe not what you're looking for but I know 2 reverbs that can do this and you can tune them. Solaris and Raum. I did this the other day with some piano notes. Raum smeared them into the most wonderful pad sound. (With a little help from Refrakt)

[CinningBao]

There are a lot of spectral/FFT solutions here, but the OP was asking for time-domain solutions.

If you want to keep in the time domain you should be looking at all-pass filters which mess with the phase of certain frequencies of the sound whih, at extreme settings, can 'smear' the sound. You could pre-filter what you're feeding to the all-pass filter to smear that particular range.

My first place to stop would probably something like SonicCharge's Echobode (with a pre-filter) or maybe Phonolith's Cascade which can create an extremely clean diffusion.

There are plenty of fourier-based solutions above so I hope you find what you're looking for!