Coastal overdrive | in the style of Strymon Riverside (high+low gain)


Hi guys,

My first attempt at re-amping a pedal without mic’ing, just using the Dwarf.
I did some manual aligning because there was a brief latency/difference on the signals.

If useful, I can do different trainings and different settings.
ESR was my lowest ever.

Strymon Riverside HIGH GAIN
presence setting on the back: middle setting

Dwarf with audio player: output 1 → input 2

Yes my foil is STILL ON! :smiley:

device = MOD-DWARF
file_name = Strymon-Riverside
unit_type = LSTM
 77% 383/500 [14:20<03:58,  2.03s/it]validation patience limit reached at epoch 384
 77% 383/500 [14:23<04:23,  2.25s/it]
finished training: Strymon-Riverside_LSTM-12
ESR after training:  0.004199963063001633

Coastal-overdrive-high-gain.json (32.0 KB)

LOW GAIN (no mid push)

device = MOD-DWARF
unit_type = LSTM
size = 16
26% 131/500 [03:50<09:20, 1.52s/it]validation patience limit reached at epoch 132
26% 131/500 [03:52<10:55, 1.78s/it]
ESR after training: 0.13942387700080872

Coastal-overdrive-low-gain.json (53.8 KB)

In an effort to get the same delay on both tracks, I looped the input signal to the output and an input again, so the path of both signals is just as much delayed.


Can you make light breakup setting with mid on and off? I had riverside and it sounds really good! Curently I switched to Kernom Ridge, IMO it is more flexible and offcourse has different sounds than riverside.

@falkTX would be great a “latency delay” for making automatic alignment at this scenario

1 Like

this is done on the training level side. The script auto aligns the input and target files when starting the training.

1 Like

I’ll try this next!

Indeed, I was thinking as well, I can imagine that the latency for this kind of setup can be predicted (and countered with a “latency plugin”?

How accurate is that? How can the script know where, in this kind of shape, where the beginning of the little tick is in the target signal? Here I manually aligned it



Update: I added a low gain (breakup point) without mid push


I was wondering, should we stick to real names or go to “pseudo’s”? like this?



@LievenDV sticking to the real name can violate trademarks. We want to avoid that completely in plugin names and artwork on our platform.

However even with registered trademarks, it’s allowed to use them in a descriptive way. So if an original piece of hardware is being used, it can be called by its name in the plugin description. To further avoid confusion, adding a disclaimer like this is recommended:

Product names and trademarks are the property of their respective holders that do not endorse and are not associated or affiliated with MOD Audio or [Plugin Author], they were used merely to identify the product whose sound was incorporated in the creation of this plugin.


can;t wait to try it. I would love to make also some prfiles since I own some drives but it is still a magic for me :slight_smile:

I thought so, thanks for clarifying that.

Would this work?

So, let’s say I call this the “Coastal Drive” overdrive and I say that is was created by modelling a Strymon Riverside and I add the disclaimer. The colors+texture are a hint but the font is different, there is a free of use image and the name is “faraway” hint.


It’s not hard but even though I’m following the same steps, it still is hit and miss for me :smiley:


If somehow I will manage to record some waves files - can anybody change them into json?

graphics look really great and I like the approach. Have a visual guide for what it is.
This could be easily a web template where you could mix and match fonts and graphics with some colors.
Load that from the .json and have it centralized.


Yeah, perhaps I should pour the format into a square and we can create some kind of catalogue?

that would be great. Can you do a separate thread for that?

Yeah let me think how we can manage that. We already have a sharing category. Perhaps a sticky thread on top of that?

People that have created model could contact me to decide on a pseudo name name and I make a fun “avatar” to go with it, some standard format to go with it (name, type, which training models, tips on use etc)

I like that and I assume pretty much it’s legally fine.


Sorry, your post got buried,

Do you still need help with profiling? It isn’t that hard it follow the guide on the page itself.
There is updated documentation on its way I’ve heard! :wink:

@Lukasz @LievenDV there is even a video guide explaining every step.

1 Like

in the early stages for sure but this will be hard to maintain when more people model their stuff.