This is an addictive workflow for generating music in the ACE-Step v1.5 model.
This workflow has an Ollama embed with complex prompting that takes any idea, theme, or random text in the "Story Input" and generates a set of style and song lyrics, which then feeds into the ACE-Step v1.5 model to produce a song of your chosen length.
There are switches for the witches that want to shut out the auto-gen and use the manual prompting system. There is a mixer and compression/clipping node in there to keep the output tidy. There's a box for LoRas. It'll all be pretty basic stuff if you've worked with Ace before, but the ACE-Step v1.5 setup will require some understanding of their git. https://github.com/ace-step/ACE-Step-1.5
I have a few Ace loras I need to upload, but you can use your own.
The Ollama node group takes a hot sec to generate lyrics, but I recommend a Qwen 8b model or a similar model.
I've left a bunch of random style notes in the notes section.
I'll do a proper write-up later. Ask your questions, and I'll try to make it all make sense in the write-up.