sema-engine is a Javascript library that provides a high-performance audio engine for modern Web applications, with an easy-to-use API. It was extracted from sema, an app developed by @frantic0, @chriskiefer and @thormagnusson, and refactored for the MIMICproject.com.

sema-engine builds upon the following components:

• the Maximilian DSP C++ library – from which sema-engine consumes DSP objects, as a git submodule

• the Web Audio API Audio Worklet – packs a bespoke Audio Worklet node (src/engine.js) and processor (maxi-processor), which loads Maximilian DSP objects and dynamic program specifications

• the Nearley compiler – generates parsers from an EBNF grammar specification

The sema-engine library exposes ES and UMD modules (works on the browser, with modern native JS modules and older JS module formats—amd, cjs for nodejs applications—think electron!).

sema-engine uses Github Actions workflows for build automation and continuous integration. The development builds propagate source maps (.map files)—so you can have modern debugging features like using breakpoints in the context of the client application. The production build ships all formats optimised and minified.

The sema-engine is published in the Node Package Manager (NPM) registry. If you are developing a Web application in a modern environment, and using a bundler such as Webpack or Rollup, you can easily add sema-engine as a dependency,

npm install sema-engine

For an advanced use, check how sema-engine integrates with Sema, a full-fledged application from which sema-engine was extracted.

You can use also use the sema-engine library modules in an a HTML file using inline <script> tags (check the published example which is output by the development build).

<script type="module">

    import {
      Engine,
      compile,
      Learner,
      getBlock
    } from "../index.mjs";

</script>

Note the that the script tag for the sema-engine module with index.mjs has type = module.

When initialising sema-engine, you need to pass the origin URL which points to where package dependencies – e.g. maxi-processor.js and maximilian.wasmmodule.js (check the dist/ folder) – should be served from.

  let engine,
      analyser = 0,
      dspCode,
      learner
      ;

  let origin = document.location.origin;

  const $ = (elemId, callback) =>
    document.getElementById(elemId).addEventListener("click", callback);

  $("playButton", "click", () => {
    engine = new Engine();
    engine.init(origin);
    engine.play();
  })

Note that the engine will make its operations depend on the origin URL for instance for loading audio samples, which should be pointed to using a relative path to the origin like so:

  $("loadSamplesButton", "click", () => {
    if(engine){
      try{
        engine.loadSample("909.wav",       "/audio/909.wav");
        engine.loadSample("909b.wav",      "/audio/909b.wav");
        engine.loadSample("909closed.wav", "/audio/909closed.wav");
        engine.loadSample("909open.wav",   "/audio/909closed.wav");
      } catch (error) {
        console.error("ERROR: Failed to compile and eval: ", error);
      }
    }
    else throw new Error('ERROR: Engine not initialized. Press Start engine first.')
  })

  $("learnerButton", "click", async () => {
    if(engine){
      try{
        learner = new Learner();
        await engine.addLearner('l1', learner);
      }catch(error){
        console.error("ERROR: Error creating or initialising learner: ", r);
      }
    }
  });
</script>

To compile the livecode, you need to do it against its grammar language specification, with the compile function. Only then you can inject the resulting code in the engine and evaluate it.

For the JS code, we provide getBlock, an utility function that pulls code from an editor block. Blocks in a Codemirror editor instance are delimited by ____ (3 or more underscores).

  const evalLiveCode = () => {
    if(engine){
      try{
        const { errors, dspCode } = compile( editorGrammar.getValue(), editorLivecode.getValue() );
        if(dspCode){
          console.info(editorLivecode.getValue());
          engine.eval(dspCode);
        }
      } catch (err) {
        console.error("ERROR: Failed to compile and eval: ", err);
      }
    }
    else throw new Error('ERROR: Engine not initialized. Please press Start Engine first.')
  }

  const evalJs = async () => {
    if(learner && editorJS){
      const code = getBlock(editorJS);
      learner.eval(code);
    }
    else throw new Error('ERROR: Learner not initialized. Please press Create Learner first.')
  }

If you would like to make contributions to the sema-engine custom WAAPI audio worklet processor, there are two compilers required to build it:

• Emscripten SDK https://emscripten.org/docs/getting_started/downloads.html

• Cheerp compiler https://leaningtech.com/cheerp/

Check the Makefile to better understand the compilation workflow.

If you are cloning this library for the first time:

1. Initialise the Maximilian and the Open303 submodules

git submodule update --init --recursive

2. Build the WAAPI Audio Worklet processor with Maximilian Pure JS, WebAssembly modules, using CHEERP and Emscripten. Build the sema-engine library into UMD and ES modules.

make

3. Build the sema-engine library into UMD and ES modules, after changes to any source file

npm run build

5. Test the sema-engine library in a local example

npm run dev

To update submodule if there are any upgrades

git submodule update --remote --merge

The sema-engine library uses Mocha for unit and integration tests.

The development build outputs the example above, which you can use to learn and test out how to work with the engine.

The sema-engine has small API surface that you can find more about on this project's wiki.

Pull requests are welcome! Fork the the project and submit PRs to the develop branch. Please observe the Contributing guidelines for other ways to contribute (issues, documentation, styles, etc).

Bernardo, F., Kiefer, C., Magnusson, T. (2020). A Signal Engine for a Live Code Language Ecosystem. Journal of Audio Engineering Society, Vol. 68, No. 1, October, DOI: https://doi.org/10.17743/jaes.2020.0016

This project has received funding from two UKRI/AHRC research grants MIMIC: Musically Intelligent Machines Interacting Creatively (Ref: AH/R002657/1) and Innovating Sema (Ref: AH/V005154/1).