A powerful multipurpose SF2/DLS/MIDI JavaScript library. It works with any modern JS environment that supports WebAssembly.

It allows you to:

• Play MIDI files using SF2/SF3/DLS files!
• Read and write MIDI files!
• Read and write SF2/SF3 files!
• Convert DLS to SF2! (and back!)
• and more!

Featuring robust MIDI analysis and editing, broadened MIDI exclusive support, and reworked, more powerful API!

Tip:

Looking for an easy-to-use WebAudioAPI browser wrapper? Try spessasynth_lib!

Project site (consider giving it a star!)

• SpessaSynth Online SF2/DLS MIDI Player
• SpessaFont Online SoundFont/DLS Editor
• signal - Online MIDI Editor

SpessaSynth Project index

• spessasynth_core (you are here) - SF2/DLS/MIDI library
• spessasynth_lib - spessasynth_core wrapper optimized for browsers and WebAudioAPI
• SpessaSynth - online/local MIDI player/editor application
• SpessaFont - online SF2/DLS editor

• Modular design: Easy integration into other projects (only load what you need)
• Flexible: It's not just a MIDI player!
• Easy to Use: Basic setup is just two lines of code!
• No dependencies: Batteries included!
• Full TypeScript definitions: Autocompletion in IDEs!

• Suitable for both real-time and offline synthesis
• Excellent SoundFont support:
  • Full Generator Support
  • Full Modulator Support: First (to my knowledge) JavaScript SoundFont synth with that feature!
  • GeneralUser-GS Compatible: See more here!
  • SoundFont3 Support: Play compressed SoundFonts!
  • Experimental SF2Pack Support: Play soundfonts compressed with BASSMIDI! (Note: only works with vorbis compression)
• Great DLS Support:
  • DLS Level 1 Support
  • DLS Level 2 Support
  • Mobile DLS Support
  • Correct articulator support: Converts articulators to both modulators and generators!
  • Tested and working with gm.dls!
  • Correct volume: Properly translated to SoundFont volume!
  • A-Law encoding support
  • Both unsigned 8-bit and signed 16-bit sample support (24-bit theoretically supported as well!)
  • Detects special articulator combinations: Such as vibratoLfoToPitch
• Sound bank manager: Stack multiple sound banks!
• Unlimited channel count: Your CPU is the limit!
• Built-in, configurable effects:
  • Reverb: Multiple characters including delay and panning delay!
  • Chorus: Modulated delay lines with multiple presets!
  • Delay: Three delay lines for all of your delay needs!
  • Insertion Effects: The ultimate effects, they can give your sounds a completely different character! (limited support)
  • GS Compatible: MIDI files can configure the effects accurately!
  • Replaceable: Effects not to your liking? You can replace them with your own!
• Excellent MIDI Standards Support:
  • MIDI Controller Support: Default supported controllers here
  • Portamento Support: Smooth note gliding!
  • Sound Controllers: Real-time filter and envelope control!
  • MIDI Tuning Standard Support: more info here
  • Full RPN and extensive NRPN support
  • SoundFont2 NRPN Support
  • AWE32 NRPN Compatibility Layer
  • Roland GS and Yamaha XG support!

• Supports MIDI formats 0, 1, and 2: note: format 2 support is experimental as it's very, very rare.
• Multi-Port MIDI support: More than 16 channels!
• Smart preloading: Only preloads the samples used in the MIDI file for smooth playback (down to key and velocity!)
• Lyrics support: Add karaoke to your program!
• Raw lyrics available: Decode in any encoding! (Kanji? No problem!)
• Loop points support: Ensures seamless loops!

• Smart name detection: Handles incorrectly formatted and non-standard track names!
• Raw name available: Decode in any encoding! (Kanji? No problem!)
• Port detection during load time: Manage ports and channels easily!
• Used channels on track: Quickly determine which channels are used!
• Key range detection: Detect the key range of the MIDI!
• Easy MIDI editing: Use helper functions to modify the song to your needs!
• Loop detection: Automatically detects loops in MIDIs (e.g., from Touhou Project)
• First note detection: Skip unnecessary silence at the start by jumping to the first note!
• Lyrics support: Both regular MIDI and .kar files!
• Write MIDI files from scratch
• Easy saving: Save with just one function!

• Level 4 compliance: Reads and writes everything!
• Compression and trimming support: Reduce a MIDI file with a 1GB sound bank to as small as 5MB!
• DLS Version support: The original legacy format with bank offset detection!
• Automatic bank shifting and validation: Every sound bank just works!
• Metadata support: Add title, artist, album name and cover and more! And of course, read them too! (In any encoding!)
• Compatible with Falcosoft MIDI Player 6!
• Easy saving: As simple as saving a MIDI file!

• Read the eXtensible Music Format files: With embedded DLS sound banks!
• MXMF support: Mobile XMF for old phone ringtones!
• Compression support: Zlib compressed files can still be read!
• Seamless integration: Read like RMIDI files!

• Easy info access: Just an object of strings!
• Smart trimming: Trim the sound bank to only include samples used in the MIDI (down to key and velocity!)
• SF3 conversion: Compress SoundFont2 files to SoundFont3 with variable quality!
• Easy saving: Also just one function!

• Same features as SoundFont2 but with now with Ogg Vorbis compression!
• Variable compression quality: You choose between file size and quality!
• Compression preserving: Avoid decompressing and recompressing uncompressed samples for minimal quality loss!
• Custom compression function: Want a different format than Vorbis? No problem!

• Read DLS (DownLoadable Sounds) files like SF2 files!
• Native support: Saving it as sf2 is still just one function!
• That's right, saving as DLS is also just one function!
• Converts articulators to both modulators and generators!
• Works with both unsigned 8-bit samples and signed 16-bit samples!
• A-Law encoding support: Sure, why not?
• Covers special generator cases: such as modLfoToPitch!
• Correct volume: looking at you, Viena and gm.sf2!
• Support built right into the synthesizer!
• Convert SF2 to DLS: limited support

• Save the MIDI file as WAV audio!
• Metadata support: Embed metadata such as title, artist, album and more!
• Cue points: Write MIDI loop points as cue points!
• Loop multiple times: Render two (or more) loops into the file for seamless transitions!
• That's right, saving as WAV is also just one function!

• Audio engine is written in pure TypeScript, so it may not be as performant as native implementations
• SF2 to DLS Conversion limits

• Improve the performance of the engine
• Potentially port the system to Emscripten

• FluidSynth - for the source code that helped implement functionality and fixes
• Polyphone - for the soundfont testing and editing tool
• Meltysynth - for the initial low-pass filter implementation
• RecordingBlogs - for detailed explanations on MIDI messages
• stbvorbis.js - for the Vorbis decoder
• fflate - for the MIT DEFLATE implementation
• tsup - for the TypeScript bundler
• foo_midi - for useful resources on XMF file format
• Falcosoft - for help with the RMIDI format
• Christian Collins - for various bug reports regarding the synthesizer
• And You! - for checking out this project. I hope you like it :)

If you like this project, consider giving it a star. It really helps out!

For use with node.js

import * as fs from "node:fs/promises";
import {
    audioToWav,
    BasicMIDI,
    SoundBankLoader,
    SpessaLog,
    SpessaSynthProcessor,
    SpessaSynthSequencer
} from "spessasynth_core";

// Process arguments
const args = process.argv.slice(2);
if (args.length !== 3) {
    console.info(
        "Usage: tsx index.ts <soundbank path> <midi path> <wav output path>"
    );
    process.exit();
}
// Read MIDI and sound bank
const sf = await fs.readFile(args[0]);
const mid = await fs.readFile(args[1]);
// Parse the MIDI and sound bank
const midi = BasicMIDI.fromArrayBuffer(mid.buffer);
const soundBank = SoundBankLoader.fromArrayBuffer(sf.buffer);

// Initialize the synthesizer
const sampleRate = 48_000;
const synth = new SpessaSynthProcessor(sampleRate, {
    eventsEnabled: false
});
synth.soundBankManager.addSoundBank(soundBank, "main");
await synth.processorInitialized;
// Enable verbose information during render
SpessaLog.setLogLevel(true, true, true);
// Enable uncapped voice count
synth.setSystemParameter("autoAllocateVoices", true);

// Initialize the sequencer
const seq = new SpessaSynthSequencer(synth);
seq.loadNewSongList([midi]);
seq.play();

// Prepare the output buffers
const sampleCount = Math.ceil(sampleRate * (midi.duration + 2));
const outLeft = new Float32Array(sampleCount);
const outRight = new Float32Array(sampleCount);
const start = performance.now();
let filledSamples = 0;
// Note: buffer size is recommended to be very small, as this is the interval between modulator updates and LFO updates
const BUFFER_SIZE = 128;
let i = 0;
const durationRounded = Math.floor(seq.midiData!.duration * 100) / 100;
while (filledSamples < sampleCount) {
    // Process sequencer
    seq.processTick();
    // Render
    const bufferSize = Math.min(BUFFER_SIZE, sampleCount - filledSamples);
    synth.process(outLeft, outRight, filledSamples, bufferSize);
    filledSamples += bufferSize;
    i++;
    // Log progress
    if (i % 100 === 0) {
        console.info(
            "Rendered",
            Math.floor(seq.currentTime * 100) / 100,
            "/",
            durationRounded
        );
    }
}
const rendered = Math.floor(performance.now() - start);
console.info(
    "Rendered in",
    rendered,
    `ms (${Math.floor(((midi.duration * 1000) / rendered) * 100) / 100}x)`
);
const wave = audioToWav([outLeft, outRight], sampleRate);
await fs.writeFile(args[2], new Uint8Array(wave));
console.info(`File written to ${args[2]}`);

Read more in the documentation

To build the NPM package, do:

npm install
npm run build

The files will be placed in the dist folder.

Copyright © 2026 Spessasus Licensed under the Apache-2.0 License.

This project is in no way endorsed or otherwise affiliated with the MIDI Manufacturers Association, Roland Corporation, Yamaha Corporation, Creative Technology Ltd. or E-mu Systems, Inc., or any other organization mentioned. SoundFont® is a registered trademark of Creative Technology Ltd. All other trademarks are the property of their respective owners.