/mecchi

Node-based web UI for AI-powered music generation.

Primary LanguageTypeScript

Mecchi

Vision

Mecchi is an attempt to build an AI-powered node-based web audio editor with the long-term goal of evolving into a full-featured modular open-source DAW designed for creating generative music content 😀

Status

But for now, it's WIP POC and not yet production-ready or close to being a mature product 🙄

Credits

It is built on Stable Audio Open, Audiocraft, AudioLDM 2, AudioSR, Bark, Flask, React, ReactFlow, Tailwind and other great projects 🔥

Short-term plans

  • Generalisation of the node UI
  • Parametric nodes
  • Config-based nodes (.py)
  • Audiogen
  • Enhanced MusicGen nodes
  • MAGNeT
  • Demucs
  • EnCLAP
  • Auffusion
  • Musicgen training
  • Lumina-T2Music
  • Decomposing SAO pipeline
  • SAO fine-tuning
  • Navigation between node groups
  • Dark theme

Pipelines

  • TTA (text-to-audio)
  • ATA (audio-to-audio)
  • ADU (audio-data-upsampling, "audio upscaling")
  • ADI (audio-data-interpolation, "audio inpainting")
  • ADC (audio-data-completion, "audio outpainting")
  • AST (audio-style-transfer)
  • ADB (audio-data-blending)
  • ADS (audio-data-separation)
  • ATT (audio-to-text, "audio interrogation")
  • MTA (midi-to-audio, "audio controlnet")
  • ITA (image-to-audio)
  • VTA (video-to-audio)
  • ATV (audio-to-video)

How it looks?

Nothing fancy yet 🐥 But here is screenshot anyway:

img

Usage

Requires Python + Git + Node.js.
To install, clone repository and run:

npm run init

To launch application, run:

npm run serve

And, if you're lucky, the app will be available at http://localhost:5555/

Try online

Open In Colab

(I am not maintaining it, to be honest. If you have issues, please let me know).

Documentation

One day 😏
OK, just need more time. Noone uses this, so it's OK to not have docs 🙂