Plus Icon Click to expand the Mega Menu

Plus Icon Click to Expand Search Input

Switch edition between U.S. Edition Asia Edition Global Edition

Variety

Plus Icon Click to expand the Mega Menu

Plus Icon Click to Expand Search Input

Variety Plus Icon Read Next: NBA 2026-2027: Here’s Where to Watch Regular Season Games Live Online

  1. Home
  2. Shopping
  3. News

Aug 20, 2026 2:46pm PT

Partner Content

AI Music Generator Mureka Launches V9.5 and O3: Music Models That (Hopefully) Sound Human

Mureka says the approach is designed to produce songs that feel more complete from beginning to end.

By Tim Chan

Plus Icon

Tim Chan

See All

AI Music Generator Mureka Launches V9.5 and O3: Music Models

Getty Images

If you purchase an independently reviewed product or service through a link on our website, Variety may receive an affiliate commission.

AI music generator Mureka is launching V9.5 and previewing the upcoming O3 model this week, which the company hopes will make it easier for aspiring artists to create music with more character and creativity.

The upcoming “O3” model takes the foundation of the company’s V9.5 AI music generation model a step further, using what it calls “test-time scaling” to give the model more room to reassess a song as it develops. Rather than committing to every early choice, it can revisit the direction of a track, adjust when an arrangement starts to crowd the vocal, or redirect the song when its structure or emotional arc begins to lose focus.

Related Stories

MILEY'S NEW YEAR'S EVE PARTY -- Pictured: (l-r) Dolly Parton and Miley Cyrus -- (Photo by: Vijat Mohindra/NBC via Getty Images)

Miley Cyrus Tears Up While Honoring Dolly Parton During Hollywood Bowl Show: 'Nothing In This Life Could Ever Tear Us Apart'

Miley Cyrus 'Bass Persuades'

Miley Cyrus' 'Bass Persuades' Has the Biggest Chart Drop Ever for an Album in the Week After a No. 1 Debut

Mureka says the approach is designed to produce songs that feel more complete from beginning to end, with sections that connect more naturally and ideas that develop with greater purpose. The aim is less about delivering an immediate “wow” moment and more about making music that holds up after repeated listens. And, despite the technology behind them, the company hopes the songs will resonate on a more “human” level as well.

Popular on Variety

Mureka currently lets its users create tracks by mood, style, purpose (I.e. film soundtrack) and more. Users can also type in lyric prompts into the generator and receive melodic inspiration to help form their song.

For its V9.5 and the upcoming O3 model, Mureka evaluated their effectiveness based on 100 prompts spanning a range of genres, with 20 evaluators reviewing the resulting tracks in blind listening tests. The company says each model was “tested under comparable prompting and generation conditions to keep the comparison consistent.”

Here’s what to know about Mureka’s latest AI-generated music models.

V9.5: Density Isn’t the Same as Quality

Among Mureka’s latest models, V9.5 posted the strongest overall results in the company’s internal testing. Its biggest change is not simply adding more layers or making tracks sound larger. Instead, the model is designed to make more selective arrangement choices, giving melodies, vocals and shifts in mood “more space to register,” according to the company.

In our testing, the difference is especially noticeable in the way songs build. Instruments tend to be introduced with more purpose, transitions feel less crowded, and the production is less likely to pile every idea into the same moment. The result is a sound that Mureka describes as closer to conventional songwriting, where “contrast and restraint can matter as much as scale.”

Vocals are also a major focus. Rather than sitting on top of the instrumental as a separate element, the lead vocal feels more connected to the arrangement around it, with phrasing, emotion and backing parts working together more naturally. In Mureka’s evaluation, V9.5 produced what listeners judged to be a convincing lead vocal in 61.0% of generations, followed prompt controls in 97.0% of cases and matched the requested genre in 95.7%. In short: it feels like one complete song, rather than multiple tracks slapped together.

Mureka is betting that stronger pacing, clearer musical decisions and more believable vocals will matter more over time than a production that tries to impress in its first few seconds.

O3: Reasoning That Looks Back

Built on V9.5, the upcoming O3 model will expand Mureka’s MusiCoT (chain of thought) framework with what the company calls “reflective test-time scaling.” In practice, that means the model can revisit decisions as a song develops — say adjusting an arrangement that overwhelms the vocal, reworking a section that loses direction, or reshaping an emotional peak that arrives too early.

Instead of committing to a single pass, O3 is designed to keep refining the track as it goes, with the goal of producing songs that feel more consistent and intentional from start to finish.

Built With the Community

With AI music generation no longer a novelty, Mureka is focusing on a harder question: can these songs earn repeat listens? The company says that goal has increasingly been shaped by feedback from its creator community.

Across more than 25,000 survey responses, users described Mureka not only as a place to make songs for fun, but as a source of inspiration, a listening destination, and part of their daily work. In interviews, creation and listening were rarely separate. The company cites music professionals who generated and auditioned ideas inside an Ableton workflow before exporting the strongest material. Others turned their own lyrics, personal milestones and stories into songs.

Long before O3 was announced, Mureka says its community had already begun asking for it. Feedback on O2 highlighted how close AI music had come to studio-quality production, while making it equally clear what users hoped to see next: stronger long-form consistency, more authentic vocals, broader genre understanding, and music that well, feels genuinely human. Those conversations directly shaped the priorities behind V9.5 and O3, leading Mureka to “optimize for realism, coherence, and replay value” rather than simply churning out louder or denser production.

Seen across Mureka’s recent releases, the progression is straightforward: V8 pushed AI music closer to release-ready production, while V9 gave creators finer control over lyrics and song structure. With V9.5 and O3, the focus shifts again, now towards music that feels natural enough, cohesive enough and distinctive enough to invite another listen. And with more and more AI-generated music content hitting streaming services, Mureka wants to make sure that its platform isn’t just creating good songs — but memorable ones as well.

Mureka V9.5 is rolling out now across the Mureka platform at mureka.ai, with O3 coming soon. 

Read More About:

Jump to Comments

More from Variety

Loading comments...

Most Popular

Must Read