FOSS AI Music is Here: Ace-Step 1.5! (Patreon)

Video - https://youtu.be/EhGt4fbQJMQ

Overview

Ace-Step 1.5 is here and with an MIT licence it is one of the best AI music generators you can run at home on your own PC.

Needs just 4GB VRAM

Makes full songs in SECONDS

ComfyUI or Gradio interface

What's not to like? :)

Ace-Step 1.5 - https://ace-step.github.io/ace-step-v1.5.github.io/

Ace-Step 1.5 GitHub - https://github.com/ace-step/ACE-Step-1.5

Ace-Step Demo - https://huggingface.co/spaces/ACE-Step/Ace-Step-v1.5

ComfyUI - https://github.com/Comfy-Org/ComfyUI

After testing the gradio web interface version is seems to have a few more features, so if you're looking for LoRA training that is the place to go. Everything is super-new though, so haven't tested making them as yet!

Workflows

There are two workflows attached - one basic version, and an AI enhanced version as well.

The AI Enhanced version version is closer to their Gradio interface in that it include AI generated song tags and lyrics:

The AI text generation uses Ollama, so you'll need to have that installed and running with the model(s) of your choice.

If you use a vision cable model (such as Gemma 3, 4b) then you can make a picture paint (up to) a thousand words!

If you only want tags, or only want lyrics, simply disconnect the relevant GET node:

Super simple, but loads of fun!

Enjoy :)