FULL STORY

Mureka V9.5: AI Music Learns Restraint

Kunlun's Mureka released the V9.5 music model, promising more natural vocals and restrained arrangements. Independent reviewers confirmed it eases AI music's overcrowded sameness.

2026-08-14 ~ 2026-08-19 · 3 episodes · 9 posts

Episode 1 · Mureka Releases V9.5 Music Model with More Natural Vocals (2026-08-14, 2 posts)

Mureka has released its V9.5 AI music model, emphasizing a more human and musical feel. The update delivers more natural vocals, arrangements that better follow user intent, and more accurate genre results.

Episode 2 · Mureka V9.5 Aims to Make AI Music Sound Human (2026-08-17, 2 posts)

Kunlun Wanwei released Mureka V9.5, an AI music model focused on sounding natural rather than loud. It aims to fix the 'fake AI music' problem with human-like melodies, harmonies and expressive vocals while keeping creator control.

Episode 3 · Hands-on: Mureka V9.5 Brings Restraint and Breathing Room to AI Music (2026-08-19, 5 posts)

On August 19, several independent creators put Mureka's newly released V9.5 music generation model to the test, and their core conclusion was consistent: it meaningfully eases the homogenized ailment of AI music being "too full, too loud, too crowded."

Confirmed

  • @FellMentKE's series of hands-on tests noted that most earlier AI music sounded stunning for the first few seconds and then stayed samey throughout, with every bar crammed with instruments and little dynamic breathing room. V9.5's biggest improvement is restraint—simple intros stay simple, verses know when to hold back, leaving the chorus genuine room to build toward an explosive peak.
  • When generated with detailed prompts (mood, genre, instrumentation), the results faithfully follow the instructions rather than merely capturing a similar "vibe."
  • Mureka is powered by its "Music Chain of Thought" (MusiCoT) technique: instead of treating audio as disconnected fragments, it plans the overall structure, emotional arc, and breathing space, so you can hear the restraint in verses and the eruption in choruses, with vocals blending naturally into the mix.
  • A comparative review shared by @eyishazyer said that after comparing dozens of AI music generators, the author believes Mureka V9.5 appears to have overcome the music homogenization problem.

Why it matters

Over-arranged, dynamically flat production has long been a glaring gap between AI music and human work. Through structured generation (MusiCoT) and improved prompt fidelity, V9.5 gives whole songs real ebb and flow and progression instead of only pleasing listeners in the opening seconds—seen as a key step toward making AI music genuinely usable.