betweentwomidnights/sa3.cpp-iplug2-demo — explained in plain English
Analysis updated 2026-05-18
Generate new music audio from a text prompt inside a DAW plugin.
Reshape or extend existing audio using an AI model without leaving the plugin.
Experiment with LoRA style add-ons to customize how the model generates audio.
Study an example of embedding an AI model directly inside a C++ audio plugin.
| betweentwomidnights/sa3.cpp-iplug2-demo | 9veedz/4leggedspiderbot | akashsingh3031/striver-sde-challenge-2023 | |
|---|---|---|---|
| Stars | 2 | 2 | 2 |
| Language | C++ | C++ | C++ |
| Last pushed | — | — | 2023-06-19 |
| Maintenance | — | — | Dormant |
| Setup difficulty | hard | hard | easy |
| Complexity | 5/5 | 4/5 | 1/5 |
| Audience | developer | vibe coder | developer |
Figures from each repo's GitHub metadata at analysis time.
Requires building a companion library (sa3.cpp) first with CUDA or Vulkan before the plugin itself compiles.
This project is a demo audio plugin that shows how to use a library called libsa3, which runs a music generating AI model called Stable Audio 3, directly inside a plugin rather than talking to it over a separate server. It can load either the medium or small-music version of the model and generate audio right inside the plugin process itself. The author describes it as an experiment and test project rather than a finished, production ready plugin, and says it may eventually be folded into a related project called gary4juce instead of being maintained on its own. The plugin has three modes: generate creates new audio purely from a text prompt, with an option to produce an exact 4, 8, or 16 bar loop timed to a set BPM. Transform reshapes audio you provide using the prompt as a guide. Continue extends a piece of source audio, using the desired total length you set. Shared controls include sliders for duration and generation steps, a seed field so a render can be reproduced later, a sampler setting, and BPM and musical key fields that get folded into the text prompt. In a DAW the BPM follows the host project's tempo automatically. The plugin also supports LoRAs, which are small add-on files that adjust how the model generates audio in a specific style. These can be imported in a few different formats, and most conversion between formats happens automatically inside the plugin without needing Python installed, except for one specific case involving raw training checkpoint files, which still requires a Python helper script provided in the related sa3.cpp repository. Building the project requires cloning this repository alongside a companion repository called sa3.cpp side by side, then compiling that companion library first using either CUDA or Vulkan, before building the plugin itself using CMake and Visual Studio on Windows. Model files are not required just to compile the plugin, since they can be downloaded later from within the plugin's interface. The full README is longer than what was shown.
An experimental audio plugin that runs the Stable Audio 3 AI music model directly inside the plugin instead of through a separate server.
Mainly C++. The stack also includes C++, iPlug2, CMake.
Not stated in the provided README content.
Setup difficulty is rated hard, with roughly 1day+ to a first successful run.
Mainly developer.
This repo across BitVibe Labs
double-check against the repo, no cap.