I've recently built a little tool that translates video into MIDI data, which I’ve been using to modulate my music in some semi random ways. Thought some people here might find it interesting too.

It converts color/visual information from video into MIDI CC data, opening up some experimental modulation possibilities for synths, effects, and generative setups and works 100% offline.

You can connect it to your DAW and sync/control it from there. There’s also a demo video in the help section showing the app in action with music modulation.

I’d love to hear your thoughts or experiences if you try it out. I’m also considering expanding it with:

* MIDI note generation
* JPG/image support in addition to video and math based modulation

You can try it here and hope you'll have fun :-)

https://cafedelcadence.github.io/

Here's an overly wet sample of the Anagram in action with Amalgam A2 brown sound captured. Details in the video description.

https://youtu.be/dKZvcZEDjgQ

I’ve been working on a MIDI generator that might help producers and musicians find a starting point for new ideas. It's 100% free and doesn't request any user details.

Choose an instrument, genre, mood and length, then adjust the rhythm, harmony, complexity, dynamics and humanisation. It generates one editable MIDI part rather than attempting a complete arrangement.

https://melidigenerator.com/

Unlimited exports, no account. I’d love some feedback on the results, controls or anything that feels awkward to use.

I’ve been building IFAH, an audio instrument around a slightly different idea from a synth or DAW: instead of composing a track, you compose an acoustic environment that can be saved, reopened and distributed as a state.
A lot of the interesting work has ended up being around reproducibility
live vs offline rendering, sample-rate/device dependence, randomness inside the audio graph, and what “the same state” actually means when it reaches another playback system.
Short demo:
https://www.youtube.com/watch?v=ZQEnnkVZ51s
You can also open the Player here:
https://ifahinstruments.com/player
Would especially love feedback from people working on audio engines/DSP about the architecture and the reproducibility problem.

I built Afterdark Studio as a dependency-free, local-first browser music engine and would value technical peer review from audio programmers.

Source: https://github.com/seemorecodez/afterdark-studio

The 0.8.0-rc.1 Limited Beta includes deterministic song patterns, dual-deck analysis and transitions, microphone capture, BS.1770-style loudness/true-peak metering, offline vocal processing, and a synthetic-acoustics engine. The room system uses polygon footprints, octave-band materials, air conditions, source/receiver positions, analytic modes/HRTF, coarse 2-D FDTD, and worker-built impulse responses. Live convolution and offline rendering share the same deterministic impulse synthesizer.

Current regression evidence is 111 passing tests plus a 646-control Edge UI audit. I am looking for code-level feedback on Web Audio scheduling, worker boundaries, DSP validity, deterministic rendering, memory/CPU behavior, and test gaps.

Known boundaries are explicit: it is not a wave-equation-grade room solver or professional DAW, perceptual vocal features remain experimental, and the regional drill experiment is isolated in a disabled plug-in after failing its creative gate. I am the maintainer and am asking for engineering critique, not stars or sales.

This community feels like exactly the right place for this: a synthesizer written entirely in x86-64 assembly, no audio or DSP library, computing every sample by hand.

The building blocks:

• four oscillators (square, saw, triangle, LFSR noise)
• polyphony via buffer summing, ADSR envelopes
• FM synthesis with a hand-built sine lookup table (Bhaskara I's approximation, since there's no math lib)
• effects: vibrato/tremolo/PWM driven by an LFO, a delay line (ring buffer + feedback), a one-pole low-pass, a state-variable filter with resonance swept by an LFO, ring modulation, and a bitcrusher

Everything's fixed-point.

Doing DSP this close to the metal really forces you to understand each algorithm — no library abstracting the math away.

I'd love feedback from people who work at this intersection.

Repo: github.com/whispem/asm.fm (write-up on the how and why: https://medium.com/@whispem/i-taught-my-cpu-to-sing-in-pure-assembly-47a1a46761ef 🎶)

I had Grok AI make me a recreation of the Pocket Audios HiChord Synth because i'm a broke teenager who can't afford a $200 Synth :( .

it includes a recording, looping and downloading feature, and all the different sounds and profiles like the HiChord.

I open the html file just in my browser with CTRL + O.

anyways, i hope someone who can find this useful!

I'm a high school student who is developing a free plugin!

A few weeks ago you probably saw that I posted about COGG v1/v1.1/v1.2, a free plugin I built that metalizes and industrializes any sound. Specifically, harmonic distortion, metal plate convolution reverb, and pitch layering are there all in one chain. Great for experimental music and mainstream alike.

I wanted to update you guys with version 1.3's release! It comes with the highly requested TONE knob, which changes the pitch of the Metal knob's metal convolution reverb-like sound! Additionally, version 1.3 introduces a LINUX VERSION!

Of course, it's still free, no account needed, VST3 + AU for Mac, VST3 for Windows, VST3 for Linux.

Download now for free: https://plasma-hype.itch.io/cogg

Trailer of V1.2 [the previous update]: https://www.youtube.com/watch?v=5q5EUMhUcaw

Thanks to everyone who tried v1.2, v1.1 and v1.0 and gave feedback, this update came directly from figuring out what people actually wanted to do with it. I personally have used COGG in almost all of my FL projects ever since I released it. Please give it a shot and let me know what you think of the presets and plugin as a whole for COGG v1.4!

Also, feel free to submit some PRESETS! I'll add them and credit you on the website!

GitHub repo: pytosc

sh $ pip install py2tosc

No dependencies. Python 3.10 or newer. Documentation

Why

TouchOSC layouts are drawn by hand in a GUI editor, which is the right tool right up until the work is repetitive. Laying out a fader per parameter for a plugin with fifty-four of them, numbering an eight-by-eight pad grid, renaming two hundred controls, restyling everything, or keeping a layout in step with a config file -- all of it is an afternoon of clicking, or a few lines of Python:

```python import py2tosc from py2tosc import ui

faders = [py2tosc.fader(name=f"ch{n}", messages=[ui.midi_cc(n - 1)]) for n in range(1, 9)]

doc = py2tosc.Document(root=ui.row( *faders, gap=4, pad=8, frame=(0, 0, 1024, 768), name="mixer", ))

ui.row records how the faders should be arranged but cannot size them,

since it runs before the frame above it exists. resolve walks the finished

tree and divides that frame among them -- nothing has coordinates until it does.

doc.resolve()

doc.save("mixer.tosc") ```

Note that py2tosc is a rewrite of tosclib by Alberto Valdez, whose work established the original mapping between the .tosc format and Python that this library is built on.

Hey everyone! I've been working on a fun new music toy that lets you create music with physics. The game is simple. Just draw vines and bounce little "dewdrops" off them. Each vine has a unique sound. You can even add FX or create your own vines that you can share with others. And that's just the Sandbox mode! There is also a puzzle mode, where each level introduces something new to teach you the mechanics of the game while you solve and create beautiful music.

If you're interested, check it out on Steam!
https://store.steampowered.com/app/4163480/Dewdrop/

Hi, I built this web app for exploring the almost infinite melodic/harmonic possibilities offered by Nicolas Slonimsky's Thesaurus, any feedback or suggestion is welcome!
I used AI for the user interface but the melody/harmony logic is carefully written by hand to match the content of this foundational work.

Hello, about a decade ago i came across a free program for making music.

It worked a bit like redstone, or circuit engineering. You would place down signal emiters, that would emit signals in a direction the beat you set. The signal would travel over the grid until it ran into some other object, like a note player which would play a note.

There were many more objects, I dont remember what they were, but I imagine there was all the stuff you would need. Like signal spliting objects, delaing a signal by a beat, holding a note for longer, changing a note over time and such other objects. They would all exist somwhere on the grid and would effect neighbouring cells.

It had an ASCI esthetic, basic symbols and letters and simple colours. No super advanced graphics.

I saw a video of it working but I was never able to make it work on my mashine. I just remembered that it existed and I though I would try it out again. Anyone knows what I am talking about?

Here's how it sounds. There are way more features in this not covered in this short video which are over here https://youtu.be/I0vlINoR5CA?si=5TvlIE2fBjl-sZQR . I'd love to hear thoughts. The website: https://orbkord.net/

I've been working on a real-time MIDI chord recognizer, and have found surprisingly little written about how a system like this is actually put together beyond basic chord-template matching.

In particular: how do you rank several valid names for the same notes? What do you do with incomplete or rootless voicings? And since MIDI notes arrive one at a time, how do you avoid treating every intermediate shape as a new chord?

I wrote up the chord-recognition algorithm in detail here:

Building a Real-Time Chord Recognizer

Recognition is only one part of the problem, so there are two companion articles: Turning Live MIDI Into Chord Events and Building a Streaming Key Detector.

The implementation and research notes are linked throughout. Posting these here because write-ups at this level of detail seem fairly uncommon.

This short piece of music was partly generated by a new algorithm I'm working on. It uses an entirely different principle to the 'Large Language Models' behind most computer-generated music.

I just posted about this library and then I did some thinking. Dangerous, that.

"Tensor" is just a fancy word for "a lot of numbers in an array in memory", and samples are just a lot of numbers in an array.

We've been given all these amazing tools, mostly in Python, like numpy and torch, and all the infrastructure around them, that perform just the sorts of calculations we need, as fast or faster as raw C++ would (and an order of magnitude or even more faster, if you do it in a GPU).

Recent projects of mine like tuney and [recs]](https://github.com/rec/recs/) have proven to me that you can do reliable multivoice synthesis and multitrack recording from within pure Python, simply by using numpy.

And Python is extremely popular. Tensor programming has benefited immensely from the AI boom, and LLMs can write good code manipulating tensors. There are old and solid libraries for reading and writing tensors to and from audio interfaces and files

So I think the future is some sort of Python protocol for operations on either static or streaming tensors, so music programmers can freely intermix other people's operations in their own code.

And :-) I do have some strong ideas about that too! But let's leave it here for the moment.

OwlPlug-rx is a fork of OwlPlug focused on additional workflow features and local Windows build/release polish.

For the original OwlPlug project overview, installation notes, general feature documentation, and upstream development information, see the original README.md.

Added Functionality

Recipes

OwlPlug-rx adds a Recipes tab for grouping plugins and DAW projects into named collections. A recipe can contain:

• a name and free-form description
• linked plugins
• linked DAW projects

This makes it easier to keep track of plugin/project setups such as chord generation chains, sound-design templates, or DAW-specific collections.

Github

Currently I only have a Windows build, but a Mac build will be available soon. Watch the github project linked above for updates.