Epic music from various artists,
Epic, Strings, Orchestral, Drums, Trumpets, Adventurous, Journey, Conflict, Marching, Chanting, etc.
Granularity Scale 20-25 seems to work best.
Description
FAQ
Comments (8)
So cool to see an Ace Step LoRA here! Hope to see more!
Feel free to checkout my other ACE-Step Lora's, I have about 4 on Civitai at the moment.
the workflow attached has no lora option?
Good point, I generally use the gradio interface bundled with the GitHub repo for ACE-Step
Can you share the details of how you trained it?
most important for me: how many "sample" song did you use? and did you annotate each one manually?
I really want to make an Hardstyle lora, but the annotation is a pain to do...
please enlighten me
Hi! For ACE-Step if its a genre and not specific artist, my datasets are only about 11-15 songs. For training a specific artist ACE-Step recommends about 50 songs for a dataset. You can follow training instructions here. https://github.com/ace-step/ACE-Step/blob/main/TRAIN_INSTRUCTION.md. You will need a GPU with 30GB of memory. I train on 5000 ada. You can checkout out my 80's citypop Lora. I uploaded the training dataset with the Lora.
@tarnished3029 ty i was more curious about the annotations
@blallo Ok I recommend checkout out the CItypop lora's dataset. The annotation for instrumental music is simple. Typically its the trigger word + instruments + mood + optional bpm.