Yuri Turov, developer of Xynthesizr and Shoom, just released Velvet Machine! This AUv3 effect promises velvety sounds, similar to a reverb. Velvet Machine has a highly customizable volume envelope, along with other controls that let you get creative with it.

Velvet Machine
Sonic texture generator / blur / non-linear reverb
*** WARNING ***
Velvet Machine is an Audio Unit v3 plugin and requires a compatible host such as GarageBand, AUM, Cubasis, Audiobus etc. The provided standalone app does NOT support the currently deprecated Inter-App Audio.
***
Velvet Machine is a CPU-intensive effect. A *minimum* of 5th Gen iPad, iPhone 6s or 1st Gen iPhone SE is required to run a single instance at maximum settings. With shorter times and/or lower densities CPU use is proportionally lower, but can still be relatively high.
***
Velvet Machine is a unique audio effect based on real-time convolution with enveloped velvet noise. Here is what it can do:
- Stretch and blur out even short sounds to lengths of up to 10 seconds
- Produce reverb-like textures with flexible volume envelope (including reverse, fading in and out, pulsating etc.)
- Work as a variable density random multi-tap delay (bordering on granular-like sounds)
PDF manual and video demos are available on developer website.
Features:
- Flexible volume envelope with arbitrary number of segments and curvature
- Effect time ranging from 100 ms to 10 s
- Response density adjustable from 1 to 2000 repeats per second
- Predelay up to 500 ms
- High cut and low cut filters for the effected signal
- Mix adjustable from 100% dry to 100% wet with no dry signal
- Mix Lock to switch presets while keeping the dry/wet mix constant
There's a demo of the presets on Yuri Turov's channel. Here's a video from Gavinski’s Tutorials!
Reader Comments 11
I suppose someone could get a similar result from pre-processing IR's (impulse responses) but this is more clever, and way more convenient.
It's not easy to get a clear definition of 'velvet noise'. I saw it described as the noise-like quality of late reflections. I heard examples on YT that sound like filtered (contoured) noise (like brown, pink, or white). I've heard some convolution reverbs that use noise-based IRs and they sound ok-- but for traditional spaces, it doesn't work very well.
I have a desktop plugin called Fog Convolver which gives me some control over the envelope of the IR. Velvet appears to give much greater control over the envelope, but not over the IR (?)
AFAIK, the name 'velvet noise' comes from a 2007 AES paper by M. Karjalainen and H. Järveläinen. In the context of their work it's clearly defined and the term seems to have caught on - I've seen it in a number of papers since.
What do you mean exactly by control over the IR in this case?
Fog Convolver allows me to choose an IR and then pre-delay, ramp on, ramp off, and ramp curves.
If Velvet uses a sample of velvet noise as the IR, then I'm assuming it doesn't allow one to choose a different IR. But it gives more flexible control over the entire envelope of the sound than Fog does.
Well, Velvet Machine is not a general purpose IR convolver, nor does it use one internally. There's no IR in the sense of a large wave file or buffer. VM is built with velvet noise in mind, taking advantage of its sparsity to perform convolution directly in time domain, as opposed to FFT-based 'fast convolution' algorithms. This approach has certain benefits:
- No latency
- Cheaper to perform smooth time and/or envelope changes.
- Comparable or better performance (I haven't run extensive tests, and I don't know how fast Thafknar is, but IIRC Velvet Machine used about 2 times less CPU than Thafknar for the same response length)
So yes, the control over the virtual IR is somewhat limited - envelope, density, LP and HP filters for the whole response and stereo width.
On the other hand, changing the density is just a turn of a knob, whereas with a general purpose IR convolver one would have to juggle a few dozen custom-generated IRs.
Yes. An important difference is that unlike a reverb, the 'reflection' density is constant and filtering is non-recursive. In a way, the fact that it can still work as a reverb is a by-product, but the goal was sound stretching/blurring.