Step Star and ACE Studio Open Source New Music Megamodel

On May 7, ACE-Step and ACE Studio officially released and open-sourced the music model - ACE-Step (Chinese name: 音跃). According to the introduction, ACE-Step has a parameter count of 3.5B, with the features of efficient multi-creation ability, strong controllability and easy expansion, and supports multi-language song generation, covering 19 languages such as Chinese, English, Japanese, Korean, Spanish, Russian and etc. Through the innovative architecture and training strategy, ACE-Step can be used to generate songs in a wide range of languages. Through innovative architecture and training strategies, ACE-Step improves the smoothness of melody and the fit of lyrics, making the generated music more complete in terms of melody and structure, and more expressive in terms of emotion. In addition, ACE-Step supports a variety of fine-tuning methods, including LoRA and ControlNet. ACE-Step is currently available on HuggingFace, GitHub, and as a demo. Huggingface: https://huggingface.co/ACE-Step/ACE-Step-v1-3.5BHuggingface Space Demo: https://huggingface.co/spaces/ACE-Step/ACE-StepGithub: https://github.com/ace-step/ACE-StepGithub Project Page: https://ace-step.github.io/