- The report describes an intermediate musical score that can be edited before audio generation.
- Consumer song generation and professional editing are separate product workflows; reported quality impressions are subjective.
From instructions to an editable score
ACE Music is using an editable musical score as an intermediate step between a user's request and generated audio, according to a November 16 hands-on report by QbitAI. The article describes a workflow in which notes, beats and chords can be represented symbolically and revised before a song is rendered.
That representation can also let a language-model agent inspect and modify the score, then call the music-generation system. The article recounts a team demonstration of such an agent workflow; this brief has not independently reproduced it.
Different tools and unresolved limits
The report distinguishes a consumer product for quickly generating songs from a professional workspace for adjusting vocals, instruments, pitch, pronunciation, emotion and arrangement. Its author tried several musical styles, background tracks and Chinese rap, offering subjective listening impressions rather than a controlled performance study.
The author found vocals better than expected but said Chinese-language songs still had room for improvement. The article also discusses a preference leaderboard; its rankings do not establish universal musical superiority. The reporting date is November 16, while the developer's underlying music-model work reportedly began in March 2025.
Sources & context
Go to the original material. Company claims remain attributed to their sources.
01


