How to make Freeverb in Puredata

Hinzugefügt: 2026-05-15

[00:00:15]Hi, in this video we will be making free verb as a pure data abstraction.

[00:00:23]Freeverb is a fairly popular reverb model. We will recreate it here with all its features, including one that has been left out in most implementations.

[00:00:35]First off, we'll start by retrieving Freeverb's original source code.

[00:00:41]Here it is a VST plugin, but we are only interested in the internals in the components folder. Inside revodc we can identify the main structure of freeverb shown here.

[00:00:59]For simplicity only the left channel is presented here. And here it is with additional information found in tuning.h.

[00:01:10]We'll first look at this component the lbcf for lowass feedback com filter. These two files allow us to find the following diagram for the filter.

[00:01:25]If you're wondering what those Z's mean, they simply represent taking the signal one sample in the past here and M samples in the past here. If you're wondering why they are this way, they have something to do with something called the Z transform, but we don't need to know that right now.

[00:01:48]The triangles here simply represent multiplication.

[00:01:54]Here we have D the damping factor and F the feedback gain.

[00:02:02]This part here is a low pass filter and here is its frequency response.

[00:02:09]As you can see, when D gets closer to one, the higher frequencies are less and less allowed through. This is to simulate the reflecting surfaces absorbing the energy of high pitched sounds.

[00:02:25]The low pass filter can be implemented on its own using a FX per object, which stands for filter expression like so.

[00:02:38]Now there are more efficient ways to implement this filter in pure data but I find this one to be the easiest.

[00:02:46]Then this part can be done with a delay line.

[00:02:51]We'll make the damping factor and feedback gain inlets so we can change them on the fly later.

[00:03:00]As for the delay, we'll make it an argument to the abstraction.

[00:03:06]Freeverb was originally written for a sample rate of 44.1 kohertz.

[00:03:12]So we hardcode it here and we convert the number of samples to milliseconds for the delay line.

[00:03:21]Now we can save this abstraction.

[00:03:26]Now let's have a look at this all pass filter.

[00:03:30]Here is how it's implemented and that gives the following diagram.

[00:03:37]And here it is implemented in pure data.

[00:03:42]Here the feedback is a constant 0.5.

[00:03:46]I think this filter is just used to add a few reflections in the reverb.

[00:03:54]Now we have all we need to put it all together using tuning.h. H we can fill in the delay values for the low pass feedback com filters and all pass filters.

[00:04:10]Here I have separated them in sub patches for left and right channels for legibility.

[00:04:18]Now all that is left to do is to add the feedback and then ping logic and initialize them like so.

[00:04:28]But wait, how about that secret feature I told you about? Well, here it is. It's the freeze mode. We can simply add it like so.

[00:04:46]And there you have it, your very own free verb inside Pure Data. I hope this video has been useful for you and thanks for watching.