https://github.com/smimram/learning-effect

Modeling guitar amplifiers using machine learning.
https://github.com/smimram/learning-effect

audio-processing machine-learning ocaml

Last synced: 17 days ago
JSON representation

Modeling guitar amplifiers using machine learning.

• Host: GitHub
• URL: https://github.com/smimram/learning-effect
• Owner: smimram
• License: gpl-3.0
• Created: 2023-08-31T16:02:11.000Z (about 1 year ago)
• Default Branch: main
• Last Pushed: 2023-09-02T11:43:44.000Z (about 1 year ago)
• Last Synced: 2024-08-05T15:06:36.458Z (3 months ago)
• Topics: audio-processing, machine-learning, ocaml
• Language: OCaml
• Homepage:
• Size: 44.9 KB
• Stars: 9
• Watchers: 3
• Forks: 0
• Open Issues: 0
• Metadata Files:
  • Readme: README.md
  • License: LICENSE

Awesome Lists containing this project

README

        
# Machine learning audio effects

Trying to learn machine learning for audio effects. This is my own take at

coding audio effects using machine learning. In OCaml.

It started with the

[LSTM](https://en.wikipedia.org/wiki/Long_short-term_memory)-based approach

described in

- [this github repository](https://github.com/Alec-Wright/Automated-GuitarAmpModelling)

- [this article](https://www.dafx.de/paper-archive/2019/DAFx2019_paper_43.pdf)

- [this blog

post](https://towardsdatascience.com/neural-networks-for-real-time-audio-stateful-lstm-b534babeae5d)

## Usage

In order to learn an effect you need both an original sound file (say

`source.wav`) and a processed sound file (say `target.wav`). You can start

learning with

```shell

leffect -s source.wav -t target.wav

```

This will learn the effect. While doing so it will also produce an `output.wav`

file (name can be changed with `-o`) resulting of the processing of the input

file by the current network. Some other useful parameters are

- `--rate`: the learning rate

- `--size`: the size of the network

- `--play`: play the output in realtime

At the end, it produces an `effect.json` file (the name can be changed with the

`--json` flag).

In order to process a file with an already learned effect (whose parameters are

stored in `effect.json`), you can simply do

```shell

leffect -i file.wav -o output.wav

```