Gong Junmin

Now we have acestep.vst3, based on acestep.cpp github.com/ace-step/acest… Let’s evolve ☆(￣▽￣)/$:*

yes!

I knew it! There is an ace-radio sound on 🔊

Train Custom ACE-Step Models Locally 1-Click Launcher for Sidestep, a training toolkit for ACE-Step audio generation models. Takes you from raw audio files to a working adapter without the friction. Now even has a GUI. Morpheus has created a launcher for it!

ACE Studio Video Composer doesn’t just generate tracks. It analyzes your scenes, cuts, and motion to create synced music & SFX on a full timeline. It’s like an AI audio assistant that builds your soundtrack and gives you fully editable clips. Download: acestudio.ai 🚀

Foundation-1 is available here huggingface.co/RoyalCities/Fo… Typical inference: • ~7 GB VRAM • ~7–8 seconds per sample on an RTX 3090 • runs entirely local Have fun!

I’ve recorded a demo on how to generate music videos/MVs using OpenClaw + ACE-Step Skill. It will automatically guide you through configuring the key. Sound on🔊

@Nate_Esparza thanks

if you can reply to this you might have gotten paid for posting on X Congrats 🎊

@m4zas24 sure

@junmingong @junmingong The only feedback I have is this maybe expansion into being able to train custom sound and narration/speech generation would be the next step forward

I think the best thing about ACE-Step 1.5 is that it wasn’t trained to please everyone. I accept that it has its own limitations. We’ve also received all kinds of constructive feedback from the community, and we’re listening, understanding, and thinking about what people really need.

All of this is good.

Some of these issues come from inference bugs, some from usage patterns, and others from the model itself — such as parameter scale, architecture, data distribution and size, annotation accuracy, and so on.

@scythe000 not yet, but there are some loras in huggingface and some in our discord channel: discord.com/channels/13692…

@junmingong Is there a repository of Loras anywhere?

Introducing Music Finetunes in ACE-Step 1.5 Generate individual vocals, instruments, or full tracks with stylistic consistency using a fine-tuned version of our Local Music model.

@ML_Prod67 yes, you can train text2vocal loras or midi to vocals control net

@junmingong If it can generate individual vocals. Can i control the melody like Ace Studio Voicebanks or Suno's cover feature which would make it an advanced RVC?

@100PercentRobot I believe it will improve iteratively. First, the community ecosystem will keep contributing, and second, we will provide better base models for fine-tuning later, which will raise the upper limit.

@junmingong Very nice! This has been available (in beta) in a couple UI's I've used, but haven't had the chance to try it yet. Has it gotten better?

We are in a great coding era, and the community is also building more user-friendly interfaces that can match or even surpass those centralized commercial UIs. github.com/koda-dernet/Si…

We will continue to launch fully open-source, even better local music generation models, democratizing and decentralizing AI-powered music creation. github.com/ace-step/ACE-S…