Mureka V9 is Kunlun Tech's flagship AI music model. It uses MusiCoT — Music Chain-of-Thought — to plan a song's structure before rendering vocals, melody and arrangement into a finished track.
The model reasons through intro, verse, chorus, bridge and outro before generating audio, giving full tracks a coherent, human-like song form.
Create from a text description, your own lyrics, or a melody/voice reference — vocals, harmony, rhythm and arrangement come out in one pass.
Supports 50+ musical styles and 10+ languages, with improved multilingual pronunciation over earlier Mureka generations.
Export separated stems and route into a DAW, so generated tracks can be remixed, mastered or dropped into a production workflow.
Mureka V9 is the flagship music-generation model from Mureka, the AI music platform built by China's Kunlun Tech (Kunlun Wanwei) and its Skywork AI division — the team that earlier shipped Mureka O1, billed as the world's first music-reasoning large model. V9 extends that reasoning approach with MusiCoT (Music Chain-of-Thought): instead of generating audio as one undifferentiated block, the model first plans the song's structure — intro, verse, chorus, bridge, transitions and outro — and then renders vocals, melody, harmony and arrangement to match that plan. The result is a full song, typically a few minutes long, produced from a single prompt in roughly a minute.
Creators can drive V9 from a plain text description, from their own lyrics, or from a melody or voice reference, and can steer output across 50+ styles and 10+ languages with stronger multilingual pronunciation than earlier generations. For production work, V9 exposes separated stems and DAW-friendly export, plus a developer API with commercial usage rights, positioning Mureka as a leading non-Western rival to tools like Suno. A July 2026 update, Mureka V9.5, layered on O3 'reflective reasoning' and an agentic composer (MuCo) for even more natural, score-worthy results.
Music Chain-of-Thought plans the full arrangement — sections, transitions and song form — before any audio is rendered, so tracks hold together across their entire length rather than drifting.
From a single prompt, V9 produces vocals, melody, harmony, rhythm and instrumental arrangement together as a coherent, publishable song, typically generated in around 60 seconds.
Start from a text style description, supply your own lyrics, or provide a melody or voice reference to guide the vocal timbre and tune the model follows.
Broad genre coverage and multilingual singing, with V9 notably improving pronunciation and delivery across non-English languages compared with prior Mureka models.
Export separated instrument and vocal stems for remixing, mixing and mastering, making generated material usable inside a professional production pipeline.
A public API at platform.mureka.ai lets developers generate songs programmatically, with commercial ownership of output included for building music into apps and services.
V9 builds on Mureka O1's music-reasoning foundation; the July 2026 V9.5 refresh adds O3 reflective reasoning and the MuCo agentic composer for more natural, emotionally varied results.
| Model name | Mureka V9 |
|---|---|
| Developer | Kunlun Tech (Kunlun Wanwei) / Skywork AI |
| Core technique | MusiCoT — Music Chain-of-Thought structural planning |
| Lineage | Follows Mureka O1 music-reasoning model; refreshed by V9.5 |
| Inputs | Text style prompt, lyrics, melody/voice reference |
|---|---|
| Output | Full song with vocals, melody, harmony and arrangement |
| Typical length | Full track (a few minutes) |
| Generation speed | ~60 seconds for a full track |
| Styles | 50+ genres |
|---|---|
| Languages | 10+ |
| Export | Full mix plus separated stems / DAW-friendly output |
| Commercial use | Commercial ownership of generated tracks |
| Platforms | Mureka web app and developer API |
|---|---|
| API | platform.mureka.ai |
| Credits | Credit-based; unused credits do not expire |
Turn a theme, mood or a set of lyrics into a complete, structured song with vocals — useful for songwriters sketching ideas or non-musicians creating from scratch.
Generate genre-specific, royalty-usable tracks for videos, ads, games and social content, tuned to a required style and language.
Produce vocal songs across 10+ languages with improved pronunciation, helping creators reach audiences beyond English-language markets.
Use stem and DAW export to pull generated vocals and instrumentals into a session, then remix, re-arrange and master into a finished production.
Developers can call the Mureka API to embed song generation into products, with commercial rights covering the output they create.
Mureka iterates quickly — V9.5 (July 2026) already supersedes V9 with O3 reasoning and an agentic composer — so exact behavior, quality and available options can shift between releases.
Some figures (exact song-length caps, credit allowances and pricing) are reported by third parties rather than confirmed in official documentation and can change; verify current limits on the Mureka platform before committing.
Vocal realism and coherence are strongest in well-represented styles and languages; niche genres, complex lyrics or less-common languages may yield less polished results.
Although commercial ownership is offered, creators are responsible for confirming licensing terms and avoiding infringing prompts or reference material for their specific use.
Reasoned, full-song AI music with MusiCoT structural planning, 50+ styles and 10+ languages is on its way to SharkFoto. Explore SharkFoto's available AI creative tools today, and check back to create with Mureka V9 when it lands.
Try Mureka V9 now