Suno
Last reviewed
Sources
33 citations
Review status
Source-backed
Revision
v8 ยท 9,062 words
Improve this article
Add missing citations, update stale details, or suggest a clearer explanation.
Suggest edit
CtrlK
Last reviewed
Sources
33 citations
Review status
Source-backed
Revision
v8 ยท 9,062 words
Add missing citations, update stale details, or suggest a clearer explanation.
Suno is a generative artificial intelligence company that develops a text-to-music platform capable of producing complete songs, including vocals, instrumentals, and lyrics, from simple text prompts. Headquartered in Cambridge, Massachusetts, the company was founded in 2022 by Mikey Shulman, Georg Kucsko, Martin Camacho, and Keenan Freyberg, all of whom previously worked at Kensho Technologies, an AI startup acquired by S&P Global.[19] Suno's web application launched publicly on December 20, 2023, alongside a partnership with Microsoft that integrated the platform into Microsoft Copilot.[4] By February 2026, Suno had surpassed 2 million paid subscribers and $300 million in annual recurring revenue, and in June 2026 it raised a $400 million Series D at a $5.4 billion valuation, more than double its $2.45 billion mark seven months earlier.[12][31]
The company is one of the central figures in a broader debate over AI training data and music copyright. On June 24, 2024, the Recording Industry Association of America filed a federal lawsuit against Suno on behalf of the major record labels, accusing the company of training its models on unlicensed sound recordings.[3] Suno responded on August 1, 2024, by admitting that it had trained on copyrighted material and arguing that the practice qualifies as fair use.[23] By May 2026, Universal Music Group and Sony Music moved to expand their claims from 560 works to 61,026 specifically identified recordings, raising theoretical statutory exposure from roughly $84 million to more than $9 billion.[32] The litigation, combined with rapid product launches and a steep funding trajectory, has made Suno a closely watched test case for how generative AI will coexist with the music business.
Suno sits at the intersection of three trends in generative AI: the maturation of audio foundation models, the migration of consumer-facing tools from research demos into commercial products, and the collision between AI companies and content rights holders. The product takes a short prose description (sometimes paired with a custom lyric block) and returns a fully mixed song with vocals, harmony, percussion, and arrangement. The output is not a synthesis of pre-recorded loops; the model generates new waveforms that mimic the textures of recorded music, including breath noises, microphone proximity, and stereo width.
The company is private and headquartered in Kendall Square in Cambridge, Massachusetts, the same biotech and AI corridor that houses Anthropic's East Coast office, Hugging Face research staff, and a cluster of Harvard and MIT spinouts. Suno's user base mixes hobbyists who treat the tool as a novelty, independent artists who use it for demos and reference tracks, and a long tail of creators making background music for YouTube, TikTok, podcasts, and games. The Microsoft Copilot integration also routes a steady flow of mainstream Windows users to the platform. CEO Mikey Shulman has framed the company's purpose as a radical lowering of barriers to creation, saying its goal is "a future where anyone can make music."[1] By early 2026, Shulman reported that roughly 90% of Suno's daily users create something rather than merely consume, a participation rate he likens to gaming rather than passive streaming.[19]
Suno's pace of releases is unusual for a company of its size. Between December 2023 and October 2025, it shipped seven distinct model versions (v2, v3, v3.5, v4, v4.5, v4.5+, v5), each of which raised either generation length, audio fidelity, vocal realism, or editing flexibility.[6] The compounding effect produced output that improved from clearly artificial in late 2023 to often indistinguishable from low-budget studio recordings by late 2025, depending on the genre.
Suno was co-founded in 2022 by four former colleagues from Kensho Technologies, a Cambridge-based AI company specializing in financial analytics.[19] Mikey Shulman, who serves as CEO, studied applied physics at Columbia University and earned a Ph.D. in physics from Harvard University, where his research focused on experimental implementations of quantum computing with solid-state spins. Before founding Suno, Shulman was the first machine learning engineer at Kensho and later became a lecturer at MIT Sloan School of Management, teaching natural language processing for finance.[19] Shulman is also a classically trained pianist and self-taught producer, a biographical detail he has repeatedly cited as motivation for the company.
Georg Kucsko earned his Ph.D. from Harvard in atomic and condensed-matter physics and led machine learning research at Kensho, where he applied transformer architectures to noisy financial text. Martin Camacho, the team's research lead, holds degrees from Harvard and worked on speech recognition before joining Kensho's machine learning group. Keenan Freyberg served as Head of Strategic Initiatives at Kensho and is credited as the operations and product lead. At Kensho, the four worked primarily on natural language processing and document understanding, but an experiment involving the transcription of earnings calls using audio AI sparked a realization: sound was significantly behind text in terms of model quality, and the prevailing audio research community was dominated by speech rather than music.[19]
The insight motivated the four to leave Kensho in 2022 and build a company focused entirely on audio generation. Their first technical artifact was not a music product. It was Bark, an open-source text-to-speech and audio model that produced surprisingly natural prosody compared to the dominant systems of the day. Bark was released on GitHub in April 2023 and became a viral hit among AI researchers and indie developers.[13] It demonstrated that the founding team could train audio models that handled non-speech sounds like laughter, sighs, and music humming. Bark also gave Suno credibility with potential investors and recruits.
The Suno name comes from the Hindi word meaning "listen." The founders have given different accounts of how it was selected, but Shulman has said in interviews that the short, vowel-rich shape sounded like a brand a music product could grow into.
Bark preceded Suno's music product by roughly eight months and remains the company's only open-source release. The model is a GPT-style transformer that turns input text directly into discrete audio tokens, which are then decoded into a waveform using a neural codec borrowed from Meta's EnCodec work. Bark supported 13 languages, could switch between them mid-utterance, and could generate non-verbal sounds such as crying, sneezing, and music in response to bracketed cues like [music] or [laughter].[13]
Bark's ability to produce song fragments was largely a happy accident. The team initially designed it for speech, but trained on a broader audio corpus that exposed the model to singing. Researchers on Twitter and Reddit quickly discovered they could coax it into producing short melodies and rhythmic chants. That behavior pointed the team toward a dedicated music model, which became the v1 prototype of Suno later in 2023.
Bark remains available under the MIT License on GitHub at github.com/suno-ai/bark and on Hugging Face.[13] A community-maintained C/C++ port called bark.cpp ships pre-built binaries for running the model on consumer hardware. By 2025 Bark had accumulated more than 35,000 GitHub stars, making it one of the most widely cloned generative audio repositories.[13]
Suno initially operated through a Discord bot beginning in mid-2023, using its v2 model. The Discord-only distribution was an unusual choice; it gave the team rapid feedback from a self-selecting cohort of music hobbyists, and it kept the product semi-private during a period when copyright questions were already simmering across the AI industry. The Discord community grew to several hundred thousand members before the public web app launched.
On December 19, 2023, the company launched its dedicated web application at suno.com and simultaneously announced a partnership with Microsoft.[4] Through this integration, Suno was available as a plug-in within Microsoft Copilot, allowing users to type prompts like "Create a pop song about adventures with your family" and receive complete AI-generated songs with lyrics, vocals, and instrumentation.[5] The Copilot integration was the first time most mainstream technology users encountered AI music generation. It coincided with Microsoft's broader push to surface third-party plug-ins through Copilot, and it gave Suno reach into Windows, Edge, and Microsoft 365 users without any direct marketing spend.[4]
The Microsoft Copilot integration brought significant early visibility to the platform and helped Suno reach over 10 million users within its first months. The Copilot relationship has remained intact through subsequent product upgrades, including the move from v3 to v4 and the introduction of Suno Studio.
Suno grew rapidly throughout 2024 and 2025. By May 2024, the platform had attracted over 10 million users.[1] The company's annual recurring revenue climbed from roughly $45 million in late 2024 to $150 million by October 2025, reaching $200 million by November 2025.[2] As of February 2026, Suno reported more than 2 million paid subscribers, $300 million in annual recurring revenue, and more than 100 million people who had used the platform.[12] The subscriber count had roughly doubled from about 1 million in November 2025, a step-up Shulman characterized as 300% year-over-year growth.[12] The company's workforce expanded from around 60 employees at founding to over 200 by late 2025, reaching 366 employees by early 2026.[19]
In late February 2026, Suno appointed Jeremy Sirota, the former CEO of Merlin (the digital licensing partner for independent labels), as its Chief Commercial Officer.[18] During his six years leading Merlin, Sirota had scaled annual revenue from $900 million to $1.8 billion.[18] The Sirota hire was widely read as a signal that Suno intended to move from adversarial relations with rights holders toward licensing partnerships, a shift that had begun with the Warner Music Group settlement three months earlier.[19]
Suno operates as a text-to-music generation platform. Users provide a text prompt describing the desired song, and the system produces a complete audio track, typically within seconds. The platform supports two primary input modes: users can write their own lyrics and specify a musical style, or they can provide a brief description and let Suno's built-in large language model generate both the lyrics and style tags automatically.
While Suno has not published a detailed technical paper on its proprietary models, the general architecture is understood to combine transformer-based models with diffusion models. Public statements from the founders, reverse engineering by researchers, and the architecture of Bark (Suno's open-source sibling) suggest the following pipeline:
The system uses neural audio codecs (similar to Meta's EnCodec or Descript Audio Codec) to compress audio into discrete tokens that the transformer model can process, and then to decompress the generated tokens back into audio waveforms at 44.1 kHz stereo. The v5 model is reported to use a substantially larger codec vocabulary and a deeper diffusion decoder than v4, which is partly responsible for the jump in vocal clarity.
Users can specify detailed style prompts that include genre tags (e.g., "indie folk," "midwest emo," "phonk," "EDM"), mood descriptors (e.g., "melancholic," "upbeat," "nostalgic"), instrumentation preferences (e.g., "acoustic guitar," "synth pads," "808 sub"), and vocal characteristics (e.g., "female vocalist," "raspy baritone," "choir"). The platform supports genre mashups, allowing combinations like "EDM + folk" or "hip-hop + jazz" to produce hybrid styles. Suno discourages prompts that name a living artist directly and rejects many such prompts at submission time, although users have found workarounds by describing voices in stylistic rather than nominal terms.
Suno also provides an automatic lyrics generation feature powered by a language model. Users can input a topic or theme, and the system generates full song lyrics with verse, chorus, and bridge structures. Users who prefer full control can write their own lyrics using a simple markup format that indicates song sections such as [Verse], [Chorus], [Bridge], [Pre-Chorus], [Outro], and [Instrumental]. The markup language has expanded over time and now includes performance directives such as [whispered], [shouted], [guitar solo], and [sad piano].
A typical Suno generation consumes credits according to a published schedule.[20] Each generation produces two stereo audio outputs of the same prompt by default, which the user can audition before saving or extending. Generations on the higher-tier models (v5, v4.5+) cost more credits than lower-tier ones, and longer outputs scale roughly linearly. Premier subscribers receive enough monthly credits (10,000) to generate hundreds of songs at maximum length.[20]
Outputs are returned as MP3 files by default, with FLAC and stems available to paying subscribers. Songs are persisted in a user library that can be made public or kept private, and public songs appear on a discovery feed where listeners can react, remix, or fork. The fork feature, available from v4 onwards, lets a user take another creator's prompt as a starting point and modify it, producing a new derivative work.
Suno has released multiple model versions since its initial launch, each bringing improvements in audio quality, vocal realism, song duration, and creative control.[6] The table below summarizes the public release history.
| Version | Release date | Max generation length | Key features |
|---|---|---|---|
| V1 | Summer 2023 | ~30 seconds | Internal proof-of-concept; not publicly released |
| V2 | Fall 2023 | 1 min 20 sec | Discord bot only; first public release |
| V3 Alpha | February 2024 | ~2 minutes | Limited public beta; first full-song generation |
| V3 | March 21, 2024 | 2 minutes | Improved audio quality; full-song generation from text prompts |
| V3.5 | May 30, 2024 | 4 minutes (first gen); 2 min per extend | Better song structure; more coherent compositions |
| V4 | November 19, 2024 | 4 minutes | Improved vocal quality; Extend, Cover, and Persona features; stems separation |
| V4.5 | May 1, 2025 | 8 minutes | Smarter style mashups; enhanced prompt adherence; Replace, Extend, Remaster, and Covers editing tools |
| V4.5+ | July 2025 | 8 minutes | Added Add Vocals and Add Instrumental production tools |
| V5 | September 23, 2025 | 8+ minutes | Studio-grade fidelity; natural, human-like vocals; 10x faster processing |
V1 was an internal proof-of-concept never released to the public. The earliest publicly available model, V2, launched in fall 2023 and was accessible through a Discord bot before Suno's website went live.[6] It could generate audio clips up to 1 minute and 20 seconds in length and represented Suno's first public demonstration of text-to-music generation. V2 was widely criticized at the time for muddled vocals and short outputs, but it remained the only generally available text-to-song model competitive with research demos from Google DeepMind and Meta.
Released on March 21, 2024, V3 was the first model designed for full-song generation and the version that propelled Suno into the mainstream technology press.[6] V3 doubled the maximum generation length to 2 minutes, offered substantially improved audio quality, and introduced cleaner genre recognition. Reviews in Rolling Stone, Wired, and TechCrunch described V3's outputs as the first AI music that could fool a casual listener in short snippets.[25] V3 supported both custom lyrics and automatic lyrics modes, making it accessible to users without songwriting experience.
V3 also coincided with Suno's first wave of viral content on social media. A handful of V3-generated tracks racked up millions of plays on TikTok and YouTube, often presented without disclosure that they were AI-generated. The resulting press coverage emphasized both the technical leap and the ethical questions about training data, foreshadowing the lawsuit that arrived three months later.
Launched on May 30, 2024, V3.5 focused on song structure and coherence.[6] It extended the maximum first generation to 4 minutes and allowed users to extend songs by 2 minutes per extension. The model produced more musically coherent compositions with better verse-chorus-bridge progressions, and it cleaned up the lyric-to-vocal alignment that had been a recurring complaint in V3. V3.5 shipped alongside the $125 million Series B announcement and helped position Suno as a serious commercial product rather than a research demo.[1]
Released on November 19, 2024, V4 brought notable improvements to vocal quality and introduced several features that became the basis of the platform's editing workflow.[7] The Extend feature allowed users to continue a song beyond its initial generation. The Cover feature let users reimagine existing Suno-generated songs in different styles, producing the same composition in a different genre or vocal style.[7] The Persona feature enabled users to create and save consistent vocal identities for reuse across multiple songs, addressing a long-standing complaint that Suno's vocals varied unpredictably between generations.[7]
V4 also introduced instrumental-only generation for users who wanted backing tracks without vocals, and rudimentary stems separation that returned a vocal track and an instrumental bed. The model became the default for both Pro and Premier subscribers, with V3.5 retained as a fallback for older songs that users wanted to extend in the original style.
V4.5 launched on May 1, 2025, for Pro and Premier subscribers.[8] It doubled the maximum first generation to 8 minutes and introduced non-destructive editing tools: Replace (swap out a section), Extend (continue the song), Remaster (improve audio quality), and Covers (reimagine in a new style).[8] The model showed smarter genre mashups and stronger prompt adherence, picking up on subtle descriptors like "leaf textures" or "melodic whistling."[8] Reviewers noted that V4.5 was the first version where unedited Suno outputs could plausibly be passed off as low-budget human productions in certain genres, particularly bedroom pop and lo-fi hip-hop.
V4.5+, released in July 2025, added production tools including Add Vocals (place a vocal performance over an existing instrumental track) and Add Instrumental (place an arrangement under an existing vocal).[6] The combination enabled a more modular workflow that resembled traditional digital audio workstations, and it set the stage for the Suno Studio launch later in the year.
Suno v5 launched on September 23, 2025, initially available to Pro and Premier subscribers and rolled out to free-tier users a few weeks later in a credit-rationed form.[9] The company described it as a "complete re-architecture" that delivered studio-grade audio fidelity and natural, human-like vocals.[9] The model captured subtle vocal details such as whispers, vibrato, growls, and grit. Individual instruments within complex arrangements became clearly distinguishable rather than blending together, and the model demonstrated noticeably better handling of dynamics across a song's structure.
Suno reported that v5 processed tracks roughly 10 times faster than previous versions while producing higher-quality output.[9] Internal benchmarks shared with reviewers showed v5 generating an 8-minute stereo track in under 30 seconds on the production infrastructure, compared to several minutes for v4.5 outputs of the same length. The model also expanded the range of supported languages and made significant strides on rap delivery, a long-standing weakness of generative music systems.
Suno's user-facing capabilities have expanded with each major release. The capability matrix in early 2026 includes:
[Verse], [Chorus], [Bridge], [Guitar solo], [Whispered] and other directives.Limits include a per-output cap of around 8 minutes, no MIDI export, no native DAW plug-in (although Suno Studio approximates one in the browser), and restrictions on prompts that name living artists. Voice cloning of real people is not supported.
Suno Studio is an advanced production environment built on top of the v5 engine, available exclusively to Premier subscribers. It provides tools that move Suno beyond simple prompt-based generation toward a more complete digital audio workstation experience. Key features include:
The introduction of Studio reflects Suno's strategy of appealing not only to casual users who want quick song creation but also to more experienced musicians and producers who need finer control over the output. Suno Studio first appeared in beta in September 2024 and entered general availability with the v5 launch in September 2025, positioned as the premium tier's flagship benefit.
Suno operates on a freemium model with three pricing tiers. Pricing has shifted several times since launch; the table below reflects the structure as of early 2026.[20]
| Plan | Monthly price | Credits | Max model | Commercial use | Key features |
|---|---|---|---|---|---|
| Free (Basic) | $0 | 50 credits/day (~10 songs) | V4.5-All | No | Daily credits (non-rolling); play and share only |
| Pro | $10/month ($8/month billed annually) | 2,500 credits/month | V5 | Yes | Priority generation; commercial licensing |
| Premier | $30/month ($24/month billed annually) | 10,000 credits/month | V5 | Yes | Suno Studio access; stems separation; early access to new features |
Free-tier users receive 50 credits per day, enough to generate approximately 10 songs.[20] These credits do not roll over. Songs created on the free tier cannot be used commercially or monetized, even if the user later upgrades to a paid plan.[20]
Pro subscribers receive 2,500 credits per month with commercial rights for songs generated while the subscription is active.[20] They gain access to the latest model (currently V5) and priority generation queues, which matter at peak hours when free-tier requests can queue for several minutes.
Premier subscribers receive 10,000 credits per month and access to Suno Studio, which includes stem separation, advanced editing tools, and early access to new features.[20] Both Pro and Premier plans offer a 20% discount when billed annually.[20]
Following the Warner Music Group partnership, Suno announced that in 2026 it would implement monthly download caps for paid users and restrict free-tier users to playing and sharing songs rather than downloading them.[19] The change responded to industry pressure to limit unlicensed audio leaving the platform, but it also drew complaints from existing subscribers who had previously been able to download an unlimited number of tracks.
Suno's partnership with Microsoft is one of the most consequential distribution deals in the consumer AI sector. Announced on December 19, 2023, the integration made Suno available as a plug-in within Microsoft Copilot, the GPT-4-powered assistant that Microsoft was then rolling out across Windows 11, Edge, and Microsoft 365.[5]
The integration works through Copilot's plug-in framework. A user types a request such as "Make me a pop song about a road trip with my dog" inside Copilot. Copilot detects the music-generation intent, hands the prompt off to Suno through a plug-in API, and returns the resulting audio file embedded in the chat interface.[5] The first song each day is free for Copilot users, and additional generations require a Suno account.
The partnership gave Suno reach into hundreds of millions of Microsoft users without direct marketing spend, and it gave Microsoft a high-engagement demo for Copilot at a moment when generative AI was still novel to most consumers.[4] The integration was reportedly arranged through informal relationships between Microsoft's Copilot product team and Suno's leadership; no equity or upfront payment was disclosed.
Microsoft has not invested directly in Suno through any disclosed round, distinguishing the relationship from Microsoft's much deeper financial entanglement with OpenAI. The Copilot integration nevertheless functions as a soft endorsement and has remained intact through later product changes, including the introduction of v5 and the Warner Music Group settlement.
Suno has raised a total of approximately $775 million across its Series B, Series C, and Series D funding rounds.[31]
| Round | Date | Amount | Lead investor(s) | Valuation | Key participants |
|---|---|---|---|---|---|
| Series B | May 21, 2024 | $125 million | Lightspeed Venture Partners | ~$500 million | Nat Friedman, Daniel Gross, Matrix, Founder Collective, Andrej Karpathy, Aaron Levie, Aravind Srinivas |
| Series C | November 2025 | $250 million | Menlo Ventures | $2.45 billion | NVentures (NVIDIA), Hallwood Media, Lightspeed, Matrix |
| Series D | June 3, 2026 | $400 million | Bond Capital | $5.4 billion | IVP, Forerunner, Union Square Ventures, Alkeon Capital Management, Quiet, Matrix, Lightspeed, Menlo Ventures, Schroders Capital |
In May 2024, Suno announced a $125 million Series B round led by Lightspeed Venture Partners.[1] The round valued the company at approximately $500 million.[24] Other participants included Nat Friedman (former CEO of GitHub), Daniel Gross, Matrix, Founder Collective, and notable angel investors from the AI and technology communities, including Andrej Karpathy, Aaron Levie (CEO of Box), and Aravind Srinivas (CEO of Perplexity AI).[1] At the time of the round, more than 10 million people had used Suno to create music, although only a small fraction were paying subscribers.[1]
The Series B closed roughly one month before the RIAA lawsuit was filed, a timing detail that has been noted in subsequent analyses of the music industry's response. Lightspeed has publicly defended its investment as a bet on generative audio as a category, with the explicit understanding that copyright questions would have to be resolved through both litigation and licensing.[17]
In November 2025, Suno closed a $250 million Series C round at a post-money valuation of $2.45 billion.[16] The round was led by Menlo Ventures, with participation from NVentures (NVIDIA's corporate venture arm), Hallwood Media (a music-industry investor that also has stakes in artist services and catalog), and returning investors Lightspeed and Matrix.[16] At the time of the raise, Suno reported approximately $200 million in annual revenue, with users creating around 7 million tracks and streaming 20 million minutes of music daily.[2]
The Series C raised eyebrows because it closed while the UMG and Sony Music lawsuits were still pending. It also valued the company at nearly five times its Series B mark, a step-up that reflected both revenue growth and the strategic value of the technology to large language model adjacent investors such as NVIDIA.[2] The deal was disclosed within days of the Warner Music Group settlement, which removed one major front of legal uncertainty.[10]
On June 3, 2026, Suno disclosed a $400 million Series D round at a $5.4 billion post-money valuation, led by Bond Capital with participation from IVP, Forerunner, Union Square Ventures, Alkeon Capital Management, and Quiet, alongside returning investors Matrix, Lightspeed, Menlo Ventures, and Schroders Capital.[31] The round more than doubled the company's $2.45 billion Series C valuation in roughly seven months and closed even as the Universal and Sony lawsuits remained active and the labels were moving to expand their claims to more than 61,000 works.[31][32] Shulman said Suno planned to release "a new model in the coming months" rooted in the Warner Music Group partnership, designed to create opportunities for artists and audiences.[31]
No Series A was ever publicly reported. The company is understood to have raised an earlier seed round in 2022 from a mix of angels and Matrix, but the size and terms were not announced.[29]
On June 24, 2024, the Recording Industry Association of America (RIAA) announced copyright infringement lawsuits against both Suno and Udio, filed on behalf of major record labels.[3] The case against Suno, Inc. was filed in the United States District Court for the District of Massachusetts. The case caption is UMG Recordings, Inc., et al. v. Suno, Inc., docket number 1:24-cv-11611, assigned to Chief Judge F. Dennis Saylor IV and referred to Magistrate Judge Paul G. Levenson.[22]
The lawsuits represented the first major test of the legality of training generative music models on commercial recordings.[27] They were filed against the two leading commercial music generators (Suno and Udio) within hours of each other, an obvious coordinated move by the RIAA designed to lay down a marker before either company could establish itself further.[14]
The plaintiffs included Sony Music Entertainment, UMG Recordings (a subsidiary of Universal Music Group), and Warner Records (a subsidiary of Warner Music Group), along with several affiliated labels including Atlantic Recording Corp., Capitol Records, Rhino Entertainment, and Arista Music.[14] The complaint alleged that Suno had engaged in mass copyright infringement by using copyrighted sound recordings to train its AI models without authorization.[21]
The lawsuit accused Suno of stream-ripping recordings from YouTube by bypassing encryption systems, and it included a list of specific recordings that the plaintiffs alleged Suno had ingested. Statutory damages of up to $150,000 per work infringed were claimed, plus an additional $2,500 per act of circumvention under the Digital Millennium Copyright Act.[14] Aggregating the listed recordings produces theoretical maximum exposure in the hundreds of millions of dollars, although such damages awards are almost never collected at the statutory cap.
The complaint also included sample Suno outputs that the plaintiffs argued were strikingly similar to specific recordings in their catalogs. Examples cited included generations that mimicked Mariah Carey's "All I Want for Christmas Is You," James Brown's "I Got You (I Feel Good)," Chuck Berry's "Johnny B. Goode," and the Temptations' "My Girl."[14] The plaintiffs treated the resemblance of Suno's outputs as circumstantial evidence that the underlying training data must have included the original recordings.
A distinctive aspect of the RIAA's complaint was the focus on the method by which Suno allegedly acquired its training data. The lawsuit argued that Suno had stream-ripped audio from YouTube videos and other streaming services, circumventing the technical protection measures that those services use to prevent unauthorized downloads.[21] Stream-ripping is independently actionable under the Digital Millennium Copyright Act, even apart from the copyright claims themselves. By foregrounding this theory, the RIAA gave itself an additional legal lever beyond the fair use debate and put Suno in the position of having to explain how it had obtained its training data.
Suno's August 2024 answer did not specifically confirm or deny the stream-ripping allegations, instead framing the entire question of how training data was acquired as legally irrelevant under the fair use doctrine.[23] That posture became a focal point of subsequent discovery.
Discovery in the case turned out to be consequential. In a motion filed in the Massachusetts court on May 21, 2026, Universal and Sony asked the court to expand the complaint from 560 identified works to 61,026 specifically identified recordings, which the labels described as "only a small fraction" of the total matches found in Suno's training data.[32] The labels said they had used Audible Magic audio-fingerprinting technology to match millions of their recordings against Suno's corpus.[32] If the expanded claims succeed, the theoretical statutory exposure would rise from roughly $84 million to more than $9 billion at the maximum per-work rate, although courts rarely award damages at the statutory cap.[32] Suno opposed the amendment, and a key summary-judgment hearing in the case was scheduled for July 2026.[32]
On August 1, 2024, Suno filed its answer in the Massachusetts case, arguing that its use of copyrighted material for training was protected under the fair use doctrine of U.S. copyright law.[23] The filing was notable because it openly acknowledged what the company had previously declined to confirm: that Suno's training corpus included copyrighted commercial recordings.[23]
The Suno answer, written by attorneys at the law firm Latham and Watkins, made several interconnected arguments:
Legal commentators were divided on the strength of the argument. Supporters drew analogies to Authors Guild v. Google (the Google Books case, decided in 2015) and Sega v. Accolade (1992), both of which allowed certain unauthorized copying as fair use when the copying enabled a transformative new use. Critics argued that music differs from text and software because the outputs of a music model directly compete in the same expressive market as the inputs, undermining the fair-use factor that considers market harm. They also pointed to the proliferation of generations that mimicked specific copyrighted works as evidence that the model had not truly transformed the underlying material.
The argument also drew attention because it was the first time a major generative AI company had formally admitted to training on copyrighted material without a license.[23] The admission, although strategic, lowered the evidentiary burden for plaintiffs on the question of whether training had occurred and shifted the dispute to whether such training was permissible.
The parties spent the late 2024 and 2025 in extensive discovery. The court approved a protective order covering Suno's training data inventory, which the company argued included sensitive commercial information about its data acquisition pipeline. Plaintiffs requested deposition of senior Suno engineers and the production of training logs that would identify the specific recordings used. Several of those discovery disputes were litigated through magistrate judge rulings, with mixed outcomes. By mid-2026, that discovery record had grown large enough for the labels to move to add more than 61,000 works to the case, and the court set a summary-judgment hearing for July 2026.[32]
Legal experts have suggested that any merits ruling in the ongoing UMG v. Suno proceedings is unlikely to be final before summer 2026 at the earliest, given the size of the discovery record and the complexity of the technical issues. The case is expected to be a candidate for interlocutory appeal regardless of how the district court rules.
On November 25, 2025, Warner Music Group and Suno announced a settlement of the litigation between them and a new licensing partnership.[10] The terms ended Warner's participation in the joint RIAA action. Under the agreement, Suno dropped its fair use defense with respect to Warner's catalog and agreed to license future training data from Warner under a negotiated royalty framework.[30] The deal allowed Suno to build new models trained on licensed Warner music, with WMG artists opting in for the use of their names, images, likenesses, voices, and compositions in AI-generated content.[15]
Artists and songwriters retained full control over whether and how their creative identity was used.[15] As part of the agreement, Suno acquired Songkick, a live music and concert-discovery platform, from Warner Music Group.[30] The Songkick acquisition gave Suno an entry point into live event discovery and a trove of fan-behavior data spanning the concert journey, and it signaled an intention to integrate AI-generated music with real-world artist activity.[30]
Financial terms were not disclosed, but industry coverage suggested the deal included an upfront payment to Warner, ongoing royalty payments tied to model output and platform revenue, and equity in Suno or rights to revenue from specific products.[30] The settlement did not affect the parallel cases brought by Sony Music and Universal Music Group, which remained active as of early 2026.[10] In a parallel development, UMG had separately settled its case against Suno's chief rival Udio on October 30, 2025, agreeing to build a jointly licensed AI music platform slated to launch in 2026.[33]
The Massachusetts case is the most prominent, but Suno faces international challenges as well. In November 2025, Koda, the Danish rights society representing songwriters and publishers, filed a lawsuit against Suno alleging unauthorized use of musical works for training. In March 2026, the Munich Regional Court held its first hearing in a case brought by GEMA, the German music rights organization, which alleged that Suno used GEMA-administered musical works without authorization for model training. A decision in the GEMA case is expected in June 2026.
The Koda and GEMA cases focus on musical works (compositions) rather than sound recordings, a distinction that matters in many European copyright systems where collective rights organizations license composition rights separately. The cases also raise the question of which national law applies when a U.S.-based platform trains on globally available data and serves users in multiple jurisdictions.
Suno operates in the growing AI music generation space alongside several notable competitors.
| Platform | Developer | Key strengths | Open source | Vocals support |
|---|---|---|---|---|
| Suno | Suno, Inc. | Full-song generation; intuitive interface; large user base | No (proprietary) | Yes |
| Udio | Udio | Granular creative control; inpainting; remixing tools | No (proprietary) | Yes |
| Stable Audio | Stability AI | Instrumental generation; claims to use licensed training data | No (proprietary) | Limited |
| MusicGen (AudioCraft) | Meta | Open-source; text and melody conditioning; research-focused | Yes (MIT License) | No |
| AIVA | AIVA Technologies | Classical and cinematic compositions; MIDI export | No (proprietary) | No |
| Boomy | Boomy Corporation | Royalty-distribution model; consumer focus | No (proprietary) | Limited |
| Soundful | Soundful Inc. | Template-based generation for content creators | No (proprietary) | Limited |
| Beatoven.ai | Beatoven Technologies | Background music for video content | No (proprietary) | No |
Udio is Suno's closest competitor and the only other defendant in the original RIAA action.[3] Udio was founded by former Google DeepMind researchers and launched its public web app in April 2024, several months after Suno's V3 release. The product is built around granular creative control, including inpainting (replacing specific sections of a song), advanced remixing tools, and what Udio calls "audio uploads" that allow users to provide reference material to steer generations.
Udio has attracted a following among professional producers and composers who want more granular control over the generation process. Reviewers have generally described Udio's outputs as more harmonically sophisticated than Suno's, particularly for jazz and classical-leaning prompts, while Suno tends to be ahead on vocal naturalness. Like Suno, Udio was sued by the RIAA in June 2024 for copyright infringement; the cases follow similar arguments and have largely tracked each other procedurally.[3] Udio's litigation diverged in late 2025, when Universal Music Group settled its case against Udio and agreed to co-develop a licensed AI music platform, leaving Suno as the major labels' primary remaining defendant.[33]
Udio also raised a $10 million seed round led by Andreessen Horowitz in early 2024 and a Series A later that year, but it has not disclosed a funding round on Suno's scale.
Stable Audio, developed by Stability AI, focuses on instrumental and ambient music generation. The platform launched in 2023 and shipped a substantially improved Stable Audio 2 model in early 2024 with longer generation lengths and improved style control. Stability claims to use licensed training datasets, such as AudioSparx, which may provide some insulation from copyright challenges. Stable Audio is well suited for background scores, advertising music, and ambient soundscapes, though its vocal generation capabilities are limited compared to Suno.
Meta's AudioCraft is an open-source framework that includes MusicGen, a single-stage autoregressive transformer for music generation trained on 20,000 hours of licensed music. MusicGen supports both text prompts and melody conditioning (where users upload a melody and the model generates music around it). Available through Hugging Face Spaces and the AudioCraft library, MusicGen is primarily aimed at developers and researchers rather than casual consumers. Because it is open-source under the MIT License, MusicGen has been forked and fine-tuned by independent developers and has become a common baseline in academic music-generation papers.
Additional competitors include Beatoven.ai (focused on background music for video content), Loudly (offering AI music for social media creators), AIVA Technologies (which generates classical and cinematic compositions with MIDI export), Boomy (a consumer-focused service with a built-in royalty distribution mechanism), and Soundful (a template-based generator aimed at content creators). Google's MusicLM and MusicFX also represent competition from large technology companies, though these have seen more limited public availability and have not been productized at Suno's scale.
A separate category of competition comes from licensed catalog services such as Soundstripe, Epidemic Sound, and Artlist, which sell unlimited subscriptions to human-composed stock music. These services compete with Suno not on generation technology but on the same use case (royalty-free music for video creators) and have historically captured the bulk of the YouTube and TikTok soundtrack market.
Suno released its first iOS app in mid-2024, followed by an Android app later that year. The mobile apps support the full prompt workflow and library playback, although certain editing features in Suno Studio were initially desktop-only. The iOS app reportedly accounted for a significant share of new user signups during 2024 and 2025, particularly among younger users who arrived through TikTok content featuring Suno-generated tracks.
In late 2025 Suno introduced Hooks, a short-form social feed feature that allows users to pair Suno-generated songs with video clips in a vertical, mobile-friendly format. The feature is designed to place music at the center of the content experience rather than treating it as background to visuals. Hooks resembles a music-first TikTok and represents Suno's most explicit move toward becoming a content destination rather than a tool.
The Persona feature, introduced in V4, allows users to capture a specific vocal identity from one song and reuse it across future generations.[7] The feature partially solved one of Suno's persistent weaknesses, namely that two generations of the same prompt could produce wildly different singers. Personas also became a flashpoint in industry discussions, since the boundary between a saved fictional persona and a thinly disguised impersonation of a real artist is not always obvious.
Suno is used across a wide range of contexts:
A controversial category of use involves attempts to imitate specific artists. Suno's prompt filters block direct mentions of living artists, but determined users have produced near-impersonations using stylistic descriptors. The company has responded with stricter filtering, but the cat-and-mouse dynamic has been ongoing.
Initial press coverage of Suno V3 in March 2024 was largely positive on the technical merits and skeptical about the legal and ethical implications. Rolling Stone described V3 as "genuinely shocking" in places, particularly for genres like blues and pop punk where the model handled vocal grit well. Wired highlighted the speed of progress from V2 to V3 and called Suno "the breakout product" of the early generative music wave.[25] Pitchfork was more skeptical, arguing that the model's outputs were impressive in seconds-long clips but fell apart over the course of a full song, particularly in lyric craft and harmonic development.[26]
Reviews of V4 in late 2024 were more uniformly enthusiastic, with multiple outlets noting that V4 generations could pass casual listening tests in short-form formats. Reviews of V4.5 and V5 in 2025 emphasized vocal realism and editing flexibility, although critics also pointed to a tendency toward sonic homogenization across genres.
Reaction within the music industry has been more polarized. Some artists, particularly independent producers and electronic musicians, have publicly embraced Suno as a sketching tool. Others, including major-label songwriters and members of the Songwriters Guild, have argued that the model fundamentally devalues the craft and that no amount of fair-use rhetoric can justify training on uncompensated work. Shulman drew sustained criticism in 2025 for remarks suggesting that making music the traditional way is "not really enjoyable" for most people; he later said he wished he had "chosen different words."[19] The November 2025 Warner Music Group settlement softened some of that opposition by introducing a licensing framework, but it has not ended industry concern.[10]
The largest ongoing concern is the question of training data. Suno's August 2024 fair-use filing admitted that the company trained on copyrighted commercial recordings, and the company has not detailed a comprehensive licensing arrangement covering its full historical training corpus.[23] The Warner settlement removes one rights holder from the unresolved set; the Sony and UMG cases remain, and by May 2026 those plaintiffs were seeking to expand their claims to more than 61,000 works after fingerprinting analysis of Suno's discovery production.[10][32] International proceedings in Denmark and Germany expand the legal exposure beyond U.S. courts.
Suno's models can produce outputs that closely resemble the style of a specific artist, even when the prompt does not name them directly. Industry advocates have argued that this capability amounts to a form of voice or style cloning that should require explicit consent and compensation. Suno has tightened prompt filtering over time and rejects many prompts that contain artist names, but the resemblance of model outputs to recognizable styles is a separate technical issue that cannot be filtered at submission time.
The ELVIS Act in Tennessee, passed in 2024, and the federal NO FAKES Act introduced in 2023, both target the unauthorized AI replication of voice and likeness. Suno has not been a named defendant in any voice-cloning case as of early 2026, but the regulatory backdrop is tightening.
AI-generated music has been implicated in streaming fraud schemes, in which bot-driven accounts stream large volumes of AI-generated tracks to harvest royalty payments. In September 2024 the U.S. Department of Justice indicted Michael Smith, a North Carolina musician, on charges that he had used AI-generated songs streamed by bot accounts to collect more than $10 million in royalties from Spotify, Apple Music, and other platforms.[28] The indictment did not name Suno as the source of the AI tracks involved; Smith was alleged to have used multiple generative music tools.[28]
Streaming services have responded with various countermeasures, including stricter detection of synthetic audio, slower payouts for accounts with anomalous listening patterns, and policy revisions that allow platforms to withhold royalties from suspected fraud. Suno itself does not distribute songs to streaming services, but the existence of platforms that allow direct upload of AI tracks (notably DistroKid) means that Suno-generated music can end up on Spotify and Apple Music through user action.
Session musicians, jingle composers, and stock-music producers have expressed concern that AI music generation will reduce demand for their work. The economic impact is hardest to measure in the bottom rungs of the music economy, where rates were already low and competition was already heavy. Industry surveys conducted by CISAC and the Berklee Institute in 2024 and 2025 reported significant declines in stock-music commissions and modest declines in advertising music budgets attributed to AI alternatives, although the share of decline directly attributable to Suno or Udio versus other forces is contested.
As with most generative AI systems, Suno's models can be used to produce satirical or misleading content that may be mistaken for real recordings of public figures. Suno has implemented filtering against explicit impersonation, but the broader capability remains, and election-period concerns have been raised in coverage by the Federal Election Commission and various state secretaries of state.
Mikey Shulman has served as chief executive officer since founding. Georg Kucsko leads research, Martin Camacho leads the model team, and Keenan Freyberg holds an operations role. The company's board includes representation from Lightspeed Venture Partners, Matrix, and Menlo Ventures following the Series C close.[16] Jeremy Sirota joined as chief commercial officer in February 2026.[18]
Suno is incorporated in Delaware as Suno, Inc., with operating headquarters at One Cambridge Center in Kendall Square, Cambridge, Massachusetts. The company maintains a smaller engineering office in San Francisco and has reported plans for a third office in Europe to support its international licensing operations following the Warner deal.