ACE-Step

ACE-Step is a local AI music model for musicians and producers who want to generate songs or revise existing audio. It accepts text descriptions and lyrics, with control over the length of the result.

• Song generation: Create vocal or instrumental tracks using genre tags, descriptive prompts or scene descriptions to guide the sound.
• Variations and passage edits: Generate another take of a song or replace a selected section while preserving the surrounding audio.
• Lyric replacement: Change words in generated tracks or uploaded audio while retaining the melody, vocal character and accompaniment; edits work on small passages because larger replacements can cause distortion.
• Vocals and production samples: Lyric2Vocal creates sung samples for demos and guide tracks, while Text2Samples produces instrument loops, sound effects and other musical material from text.
• Multilingual singing: Generate vocals in languages including English, Chinese, Japanese and Spanish, with weaker results in less common languages.

The model runs on your own hardware, including NVIDIA GPUs and Apple Silicon, and uses the Apache 2.0 license. Output quality varies with the random seed and requested duration. Some genres have weaker style adherence, vocals can lack nuance, and edits or extensions can introduce unnatural transitions.