Forget typing prompts. Just draw.

OBSIDIAN Neural now has a canvas where you literally draw your sound.

Sketch chaotic lines → aggressive distorted rhythms
Draw smooth curves → ambient pads
Geometric shapes → structured sequences

A Vision Language Model interprets your drawing and translates it into audio generation prompts. Takes 10-20 seconds to generate your sample, then you can loop it, layer it, perform with it.

Draw. Generate. Layer. Loop. Perform.

Links:
Website | GitHub

Open source. Cross-platform. Built for experimentation.

 

The goal of this superfluous project is, as always, cash:

 

You can build your own server with the open source code on github man or build your own server by yourself and use any models you want.
Yes I propose a paid solution, but if you want a free solution, everything is free and open source on Github.

 

Hello @innermost, thanks for the information; I'm not interested in your product.

 

"dont feed the credits monster!" - unknown.

 

...especially when the witches had nearly destroyed him

 

It's primarily the marketing strategists who are paid to mislead you.

The Pro version is €14.99 – why not round up to €15? Psychology...!
The Studio version is €29.99 – why not round up to €30? Psychology...!
Then, in the fine print, it's €59.99 – why not €60? Psychology...!

 

@PulseWave Appreciate the feedback! Just to clarify: the entire codebase is AGPL-3 licensed on GitHub - anyone can fork it, modify it, or self-host for free. The paid tiers simply cover cloud GPU costs for those who prefer convenience over DIY setup. Different approaches for different needs

 

Thanks @innermost for the interesting information.

 

So how did that sound like? Have you tried this drawing in Obsidian Neural?

No problem

Well I personaly have

 

Last edited by a moderator: Dec 2, 2025