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

[osiris]

You can always try Zyklops. I had a lot of fun with this. You can load a wave and put it to an exact grain of the wave.

[vurt]

isnt smearing in the time domain, reverb?

[mjolnir]

isnt smearing in the time domain, reverb?

yes! this, (or time expansion)!... or granular. take your pick or combine or....

[FilterEverything]

How about using your favorite convolution engine with filtered and volume shaped noise as the impulse? Filtering the noise gives exact control of what frequency range you want to "time-smear", the envelope/ADSR of the noise gives control over how you want the smearing to evolve over time, and the wet/dry mix could determine the amount of smearing.

[CinningBao]

Just remembered I did this brilliantly under-viewed video about smearing and blurring noises..

the technique FilterEverything mentioned is covered in chapter 3 at 13:33 - all the 3's, and a 1

[David Kissam]

Thanks for the video, CinningBao. The last method using PaulXStretch seems to be what I was imagining. Reverb is not suitable because it happens only after an impulse; I was looking for something that happens equally before and after an impulse, which seems to be what this is. Also, it appears to be using the method I had imagined - using large FFT blocks to blur the sound in the way that a spectrogram is blurred visually by using large FFT sizes to create it. Groovy.

[BertKoor]

I am working on samples for a character piano VI that I am making.

If this is recorded from a real acoustic instrument, then it can be fixed during the recording, eg by placing the mic further away.
Probably much too late for that now

[FilterEverything]

Reverb is not suitable because it happens only after an impulse; I was looking for something that happens equally before and after an impulse, which seems to be what this is

If you load a symmetric impulse into the convolver, you will get equal amounts of blurring before and after the transient.

[David Kissam]

If you load a symmetric impulse into the convolver, you will get equal amounts of blurring before and after the transient.

Cool, I did not know that.

[David Kissam]

I am working on samples for a character piano VI that I am making.

If this is recorded from a real acoustic instrument, then it can be fixed during the recording, eg by placing the mic further away.
Probably much too late for that now

It is indeed too late, but also I think there would have been no way to capture the tone I wanted while also avoiding the noises. A truly clever person might have sampled a different piano.

[cthonophonic]

Since I didn't see it mentioned here yet, Velvet Machine by Phonolyth does basically what you're asking for, I think:

https://phonolyth.com/products/velvet-machine

The underlying technique is basically convolution with velvet noise using an envelope of your choosing, up to 10 seconds long. It's pretty neat!

[CinningBao]

Thanks for the video, CinningBao. The last method using PaulXStretch seems to be what I was imagining. Reverb is not suitable because it happens only after an impulse; I was looking for something that happens equally before and after an impulse, which seems to be what this is. Also, it appears to be using the method I had imagined - using large FFT blocks to blur the sound in the way that a spectrogram is blurred visually by using large FFT sizes to create it. Groovy.

My pleasure David; I'm glad someone finally got some use from it!
I think the confusion (at least for me) was that you were asking for a solution in the time domain (where we'd use delays and allpass-filters to repeat and blur the sound) so I was keeping away from convolution and FFT majick.

I was briefly involved in the development of the Xenakios 'remix' of paulstretch (and I have a deepdive into it if you fancy some super-nerdery) where I managed to convince him to increase the maximum block size to around 1 minute. THAT'S AN FFT BLOCK SIZE OF ONE MINUTE. Enjoy