Make 8-bit sound effects in your browser: coins, jumps, lasers, explosions. Mash a pad until you like what you hear, then download the .wav. No account, no upload, no plugin. Every sound is arithmetic, generated in the tab.
The browser half of blip8, which is the Python library underneath, and blip8 sounds, which is the ready made pack for people in a hurry.
It exists because every game I built started by rewriting the same throwaway
sound script (the longer version).
This is that library ported to WebAudio, and the port is not taken on trust:
npm test runs every recipe the tool prints through the real blip8 and diffs
the samples.
- 🏠 blip8.sindriax.dev: the front door, all three in one place
- 🦇 blip8: the Python chiptune synthesis library.
pip install blip8 - 🧪 blip8 lab: make 8-bit sounds in your browser (source)
- 📦 blip8 sounds: free CC0 chiptune SFX packs, generated from code (source)
npm install
npm run devMash a pad, or press 1 to 7. m mutates the current sound, space replays it.
The whole app is a few hundred lines of vanilla TypeScript, no framework, so it loads instantly inside an itch.io iframe.
A sound is a Params object. Flat, plain data: a category, a seed, a voice,
eight numbers and a crunch flag. Everything in the app is a function of it.
randomize(category, seed) -> Params -> render(Params) -> Float32Array
Those two halves stay apart on purpose. randomize holds all the taste (a jump
is a rising glide at a thin duty, a hurt is a falling one that lands on a bass
note) and render knows nothing about randomness, so the sliders can hand it
anything. The ranges come from the same recipes that generated the sound pack.
Every number lands on its slider step. That one rule buys three things: a slider does not jump when you first touch it, the printed recipe is the sound rather than a rounding of it, and a shared link stays short.
src/synth.ts is a port of blip8's core to WebAudio: square, triangle and
noise oscillators with accumulated phase for pitch glides, an ADSR envelope,
wavetables, 4 bit crunch, and 16 bit .wav export.
A sound straight off a pad is only ever a category and a seed, so its link says exactly that. Once a knob moves, the roll no longer describes the sound, so the link carries every number instead.
| Hash | Means |
|---|---|
#coin.481920371 |
the coin that seed rolls |
#coin.481920371! |
the same one, 4 bit crunched |
#laser.5~sine~f1800_g0.14_... |
a laser somebody edited |
Unparseable hashes are ignored, out of range numbers are clamped rather than refused, so a hand edited link still makes a sound.
The lab prints the blip8 call for whatever you are hearing, so you can install
the library and render the same sound from Python. That is only worth anything
if it is true, so it is a test rather than a promise: tests/port-fidelity.test.ts
renders each sound in TypeScript, runs the printed recipe through the real
blip8, and diffs the samples.
npm testAgreement is within one int16 step, which is the resolution of a .wav file.
Writing that test found three real divergences in the port: the glide
interpolation and the envelope ramps were each off by one sample, because
numpy's linspace includes its endpoint, and crunch rounded ties up where
numpy rounds them to even.
Noise is the one exception, and the recipe says so in a comment. blip8 draws from numpy's PCG64 and the lab from mulberry32, so the same seed gives the same kind of noise, not the identical noise. The fidelity test skips noise for that reason, and skips itself entirely if uv is not installed.
MIT, see LICENSE. The sound pack is CC0 separately. Echo the bat (the character art) is all rights reserved: use the code, not the mascot.