Beyond Suno: Who Is Building the Next Major AI Music Platform?

Not long ago, comparing AI music platforms was relatively simple.

Which one could turn a text prompt into the better song?

In 2026, that question is becoming less useful.

Producing a complete piece of music from a prompt is no longer enough to define leadership in this market. The more important competition is moving elsewhere: toward systems in which a musician can bring in original material, preserve a recognisable voice or sonic identity, change individual sections without destroying the rest of a composition, work with stems, move material into another DAW, document its origin, and understand what rights accompany it when it leaves the platform.

In other words, the market is beginning to move from generation toward production.

That distinction changes how the major players should be evaluated.

Suno currently has perhaps the clearest example of this transition.

In June 2026 the company announced that it had raised more than $400 million in Series D funding at a $5.4 billion post-money valuation. It also said that its first music model developed in partnership with the music industry would begin rolling out in the following months.

Official source — Suno: The Next Chapter / Series D announcement

The scale is significant, but the more interesting development is happening inside the product itself.

Suno has spent the last year moving beyond the simple prompt-to-song model. Studio, improved stem separation, Voices, Custom Models, editing tools and increasingly detailed control over musical material all point toward a broader production environment. Suno itself now describes Studio as its professional production suite.

The September 2026 changes make that direction even more visible.

Pro accounts now have 20 standard downloads per month and Premier accounts 60, while Premier users working through Suno Studio retain unlimited Studio exports. Suno explicitly says that Studio is intended for generating and manipulating stems and samples.

Official source — Suno: September downloads policy and Terms update

As discussed in our previous paper, this creates an important distinction between mass export and professional production.

The value of the highest tier is no longer simply the ability to generate more material. It is increasingly the ability to take a piece of material seriously enough to work on it.

The new Terms, effective 3 September 2026, also make the responsibility surrounding uploaded material more explicit. Rights to inputs, permissions and the legal chain surrounding source material therefore become increasingly relevant to professional work.

Official source — Suno: Terms of Service

This is where Suno’s current advantage is strongest.

Not necessarily because every individual generation will be better than every result produced elsewhere.

But because a production ecosystem is beginning to form around the model.

If there is one company whose development deserves particularly close attention today, it is ElevenLabs.

Until recently, ElevenLabs was still widely associated primarily with synthetic voice technology.

That description is now badly outdated.

In May 2026 the company introduced Music v2, improving vocals, instrumentation, arrangement and multilingual performance. More importantly, the model introduced granular inpainting and section-by-section composition rather than requiring the user to regenerate an entire song whenever something needed to change.

Official source — ElevenLabs: Introducing Music v2

Then the platform began expanding rapidly.

In July, ElevenLabs introduced References. A creator can upload an existing track and use it to guide the style, instrumentation and overall feel of a new generation. ElevenLabs says the reference material is checked for copyrighted recordings belonging to others before it can be used, making the feature explicitly oriented toward music the creator is authorised to provide.

Official source — ElevenLabs: References for Music v2

The company also introduced Vocals, allowing a reusable vocal identity to remain consistent across different songs. A user can build a Vocal from their own recordings, while Styles provide a related mechanism for preserving musical identity.

Official source — ElevenLabs: Vocals and consistent singing identity

Even more interesting from a professional perspective is Music Finetunes.

ElevenLabs allows a model to be fine-tuned on material supplied by the creator so that future generations retain a particular sonic identity. Its API version creates private Finetunes available only to the user who trained them. The ElevenCreative version can generate vocals, instruments or complete tracks aligned with the submitted material.

Official source — ElevenLabs: Music Finetunes

And on 25 August, ElevenLabs introduced Composer.

Composer allows individual sections to be regenerated independently. A verse can be rewritten, a bridge slowed down, or the key of a chorus changed without replacing the rest of the song. Multiple alternatives can be generated before one is selected.

Official source — ElevenLabs: Composer — section-by-section song editing

That is a substantial shift.

The less often a musician has to tell a system “make the whole song again,” the closer that system becomes to a genuine production tool.

ElevenLabs is also building its music business explicitly alongside parts of the established music industry. Its work includes partnerships with Kobalt and Merlin, while ElevenMusic itself combines creation, remixing, publishing and artist monetisation.

Official source — ElevenLabs: Introducing the Eleven Album

There is, however, an important practical distinction around rights.

Eleven Music’s commercial permissions depend on the subscription plan. Its current model-specific terms distinguish between download rights, streaming rights and other media uses; streaming distribution is prohibited on the lowest plans and permitted on higher qualifying plans.

Official source — ElevenLabs: Eleven Music model-specific commercial terms

That detail matters.

“Commercial use” is not always one universal permission. Different platforms increasingly divide it into different forms of exploitation.

For professional artists, those distinctions need to become part of the production decision itself.

Another serious participant in this market receives considerably less attention in Western creative circles.

Mureka has been developing rapidly, and its current product is already much more than a simple song generator.

On 28 August 2026, Mureka released V9.5, with improvements aimed at more natural song structure, more convincing vocals, greater space in arrangements and more stable prompt control.

Official source — Mureka: What Changed in V9.5?

But again, the model itself is only part of the story.

Mureka Studio supports timeline-based work, extension of tracks, adding instrumental or vocal material, and stem extraction. Its current tools can separate a composition into as many as 12 tracks and export separated material as high-quality audio or MIDI for work in Logic, Ableton, FL Studio and other DAWs.

Official source — Mureka: Studio features and 12-stem workflow

Its Premier offering includes Studio, reference-based creation, remixing, 12-stem separation and MIDI/WAV export.

Architecturally, that puts Mureka in the same broader race as Suno and ElevenLabs:

not simply generate a result, but continue working on the result.

Mureka’s paid API also grants commercial authorisation for generated material, including platform distribution, advertising and other business uses.

Official source — Mureka API: commercial-use FAQ

That does not automatically answer questions of copyright authorship or legal protectability. And that distinction is becoming one of the defining issues of AI-assisted music:

permission to commercially use an output and copyright protection for that output are not the same thing.

Udio was for a long time the most obvious direct rival to Suno.

Today it occupies a much stranger position.

In October 2025 the company entered into a partnership with Universal Music Group. In November it announced a licensing arrangement with Warner Music Group as well. The Warner agreement describes a future service in which users will be able to create remixes, covers and new songs using the voices and styles of participating artists, with new revenue streams and safeguards for the artists and songwriters involved.

Official source — Udio: Warner Music Group partnership

This makes Udio an important experiment.

Instead of treating licensing as something that happens only after a generative system has already been built, the company is attempting to incorporate licensed participation by rights holders and artists into the architecture of the next system.

But the transition has created a major practical limitation.

Udio currently states that downloading audio, video and stems has been disabled.

Official source — Udio: changes associated with the UMG partnership

For a casual creator, that is inconvenient.

For a professional workflow that depends on moving material into other software, archiving sources and continuing production outside the generator, it is much more significant.

For the moment, Udio may therefore be more important to watch as a prototype of a licensed AI music ecosystem than as the central production tool for an independent artist.

Google does not currently offer a simple copy of Suno.

That may be precisely why Lyria should not be underestimated.

On 29 July 2026 Google introduced Lyria 3.5 in Flow Music, with improvements to musicality, lyric generation, vocals and pronunciation, together with more direct control over tempo and duration.

Official source — Google: Introducing Lyria 3.5 in Flow Music

Taken only as a current music-production interface, Lyria does not yet represent a complete replacement for systems such as Suno Studio.

But the context is different.

Lyria sits within an ecosystem that includes Google DeepMind, Gemini, Google’s cloud infrastructure and YouTube.

Google therefore does not necessarily need to win today’s competition between standalone AI song generators.

It has the scale to influence where AI music is eventually created, distributed, discovered and connected to other forms of media.

That is potentially a much larger competition.

ByteDance presents another unusual case.

Its Seed-Music research is not currently packaged as a mass-market Suno replacement, but the technical direction is highly relevant.

Seed-Music demonstrated music generation conditioned on multimodal inputs, note-level editing and zero-shot singing voice conversion from approximately ten seconds of a user’s singing or speech.

Official source — ByteDance Seed: Seed-Music

Its audio prompting work separates two particularly interesting concepts.

Continuation preserves elements such as melody and rhythm so that a generation can behave like another part of an existing work.

Style Transfer creates different melodies and chords while retaining the broader aesthetic environment of the reference.

Official source — ByteDance Seed: Seed-Music Audio Prompting

ByteDance is also expanding beyond music itself.

In July 2026 it released Seed Audio 1.0, a system intended to generate speech, ambience, sound effects and other elements of an audio scene within a unified framework. The company says future development will extend fine-grained control to additional sound elements, including music and multitrack audio generation.

Official source — ByteDance Seed: Seed Audio 1.0

That wider context matters.

ByteDance is simultaneously developing major systems for video, image, voice and audio creation.

Its long-term advantage may therefore not come from creating another isolated music generator.

It may come from combining music, voice, image and moving picture into a single production ecosystem.

Stable Audio is interesting for almost the opposite reason.

Stable Audio 3.0 is not currently a direct replacement for Suno in vocal songwriting. Stability itself notes that the system does not generate vocals or lyrics.

But the company is placing unusual emphasis on different values: licensed training data, commercial output rights, downloadable model weights and the possibility of running models independently of a central platform.

Stable Audio 3.0 includes open-weight Small, Medium and Small SFX models, while the Large model is available through API and enterprise self-hosting. Stability says the family was trained on fully licensed data and permits commercialisation of outputs under the applicable licence.

Official source — Stability AI: Stable Audio 3.0

This represents a different possible future.

For many musicians, the most attractive platform today may be the one that produces the best vocal performance.

For professional infrastructure, however, another question may eventually become just as important:

Who controls the model, the data, the production environment and access to the work once it has been created?

A downloadable or self-hosted model approaches that question very differently from a cloud platform that can change export rules overnight.

If the only criterion is the quality of a finished song generated from a prompt, any ranking quickly becomes subjective.

A professional workflow produces a different picture.

At the beginning of September 2026, Suno appears to have the most mature all-in-one environment for creating a complete song and then continuing to work on its components.

ElevenLabs is perhaps the fastest-moving direct challenger, particularly around persistent vocal identity, reference audio, private musical finetuning and section-level editing.

Mureka already has a surprisingly deep production environment and should no longer be treated as a peripheral alternative.

Udio is conducting one of the industry’s most important experiments in licensed generative music, although its current export limitations make it difficult to position as the main professional tool.

Google Lyria and ByteDance Seed represent something different again: companies large enough that they do not need to imitate Suno in order to reshape the market.

And Stability AI is testing another direction entirely — one in which licensing, portability and control over the underlying model may become part of the definition of professional independence.

The first stage of AI music was built around surprise.

Write a few lines.

Wait a moment.

Hear a song.

The next stage will be built around control.

What did the human creator bring into the process?

Can the original melody be preserved?

Can a musician use their own performance as a source?

Can a singer maintain the same vocal identity across multiple works?

Can one instrument be replaced without recreating everything else?

Can a single verse be changed while an existing chorus remains untouched?

Can the creator obtain stems?

Can the origin of the material be documented?

Can the work continue outside the platform?

What rights accompany the result once it reaches Spotify, film, advertising or another commercial environment?

And perhaps most importantly:

can we still see the human creative contribution after the technology has finished its part of the work?

These questions are beginning to separate an AI generator from an AI production instrument.

That is why the future competitor to Suno may not be the company that produces a slightly more impressive voice next month.

It will be the system that gives a musician more control over the relationship between human source, machine operation and final work.

The direction taken by Suno, ElevenLabs, Mureka, Udio, Google, ByteDance and Stability AI suggests that this race has already begun.

And if the current trajectory continues, the central question of AI music may soon change again.

Not:

“Was AI used?”

but:

“What did the artist do — and what did the tool do?”

That is a much more useful question.

And it is also the question around which a professional AI-assisted music industry may eventually have to organise itself.

---

From Generation to Production: What Suno’s New Rules Mean for Professional Artists

The previous paper examines Suno’s September 2026 download rules, professional Studio workflow, source documentation and the need to describe the specific role performed by AI rather than reducing a complex production process to the label “AI-generated.”

• Suno: Series D announcement · September downloads policy and Terms update · Terms of Service
• ElevenLabs: Music v2 · References · Vocals · Music Finetunes · Composer · Music Model-Specific Terms
• Mureka: V9.5 release notes · Studio features · Commercial-use FAQ
• Udio: Warner Music Group partnership · UMG transition and download changes
• Google: Lyria 3.5
• ByteDance: Seed-Music · Audio Prompting · Seed Audio 1.0
• Stability AI: Stable Audio 3.0