Udio
Last reviewed
May 13, 2026
Sources
62 citations
Review status
Source-backed
Revision
v4 ยท 9,551 words
Improve this article
Add missing citations, update stale details, or suggest a clearer explanation.
Suggest edit
CtrlK
Last reviewed
May 13, 2026
Sources
62 citations
Review status
Source-backed
Revision
v4 ยท 9,551 words
Add missing citations, update stale details, or suggest a clearer explanation.
Udio is an artificial intelligence music generation platform developed by Uncharted Labs, Inc. that creates full songs from text prompts. Founded in December 2023 by a team of former Google DeepMind researchers, the platform launched its public beta on April 10, 2024, and quickly became one of the two leading AI music generators alongside Suno. Udio allows users to produce complete, production-ready tracks with vocals, instrumentation, and lyrics by describing the desired style, genre, mood, and lyrical content in a text prompt.
The platform gained widespread attention in mid-2024 after a comedic AI-generated track called "BBL Drizzy," created by a user on Udio, went viral and was later sampled in a commercially released song by Drake collaborator Sexyy Red and producer Metro Boomin. Within weeks of its launch Udio also became the subject of one of the most consequential copyright lawsuits in the generative AI era, when the three major record labels (Universal Music Group, Sony Music Entertainment, and Warner Records) sued Uncharted Labs in the Southern District of New York. By late 2025 Udio had settled with both UMG and WMG and pivoted toward a fully licensed AI music creation platform planned for 2026, while Sony Music's case continued in federal court.
Udio is built and operated by Uncharted Labs, Inc., a Delaware corporation with its principal office in New York City and an additional office in London. The product offers a simple browser-based interface where users type a description of the song they want (genre, decade, mood, instruments, vocal style, subject matter) along with optional lyrics, and the system returns a fully mixed and mastered audio clip with vocals and instrumentation. Generations are short by default, typically 33 seconds per take, and longer compositions are built by extending those takes forward and backward up to a 15 minute ceiling.
The service competes most directly with Suno, the other major consumer AI music generator. Where Suno has been praised for catchy, radio-ready short songs, Udio has built its reputation on higher-fidelity 48 kHz stereo output, more polished instrumental mixes, and editing tools that resemble a digital audio workstation more than a chat interface. The two companies have followed remarkably parallel paths since the spring of 2024: parallel launches, parallel funding rounds, parallel lawsuits filed on the same day by the RIAA, parallel settlement negotiations with the three majors, and parallel transitions toward licensed catalogs and walled-garden platforms.
This article covers the founding team, funding history, product capabilities, model versions, pricing, the copyright litigation and subsequent settlements, viral cultural moments like "BBL Drizzy," comparisons to Suno, and the implications for music creation and copyright law going forward.
Udio was created in December 2023 under the corporate name Uncharted Labs, Inc., a Delaware corporation headquartered at 750 Lexington Avenue, Floor 9, New York, New York 10022. The company also maintains an office in London, where several of the founders were based during their tenure at Google DeepMind.
The five co-founders are all former research scientists or engineers from DeepMind's audio and generative modeling groups. Their work at DeepMind included contributions to projects such as Imagen (the text-to-image model) and Lyria (Google's music generation model). CEO David Ding has stated in interviews that he previously led a 30-person research team at DeepMind, where the group focused on diffusion and likelihood-based models for media generation.
| Co-founder | Role at Udio | Background |
|---|---|---|
| David Ding | Co-founder, CEO | Former Google DeepMind researcher; led a 30-person team working on generative models for audio, image, and video |
| Conor Durkan | Co-founder | Former DeepMind research scientist in London; PhD from the University of Edinburgh in generative modeling and likelihood-free inference; worked on Imagen and Lyria |
| Charlie Nash | Co-founder | Former DeepMind researcher; published work on autoregressive and diffusion-based generative models |
| Yaroslav Ganin | Co-founder | Former DeepMind researcher; previously known for foundational work on domain-adversarial training of neural networks and on differentiable rendering |
| Andrew Sanchez | Co-founder, COO | Former DeepMind researcher; left to handle product and operations |
At the company's public launch in April 2024, Ding framed Udio as the culmination of years of generative modeling research finally being directed at music specifically rather than as a side application of broader audio work. "There is nothing available that comes close to the ease of use, voice quality and musicality of what we've achieved with Udio," he said in the company's announcement materials. In an early podcast interview with Music Ally, Ding and Sanchez emphasized that the goal was less to replace musicians than to give non-musicians a way to participate in song creation. "We want to give people a chance to showcase their creations and to have other people respond to it," Ding said.
The founding team's prior DeepMind work proved relevant in two ways. Technically, the architectures and training methods used at DeepMind for image and audio generation (especially diffusion-based approaches and large transformer backbones) carry over directly to music synthesis. Culturally, the team had spent years working on generative models inside a research lab that emphasized safety reviews and dataset documentation, which became newly important after the RIAA filed suit in June 2024 and dataset provenance became a central legal question.
Udio operated in a closed beta from late 2023 into early 2024, with invited testers including musicians, producers, and tech industry insiders. The public beta launched at udio.com on April 10, 2024, with a simultaneous press release coordinated through PR Newswire and a flurry of coverage in Rolling Stone, The Verge, VentureBeat, Music Ally, and Music Business Worldwide.
The launch product was deceptively simple. Users entered a text prompt and Udio returned two roughly 33 second song candidates, each fully mixed and mastered with vocals and instrumentation. Prompts could specify genre, mood, decade, instruments, lyrical content, and stylistic references. Lyrics could be supplied by the user directly or generated automatically from a topic. Each free user received a daily allocation of credits sufficient to generate dozens of songs per day, and a built-in social feed allowed users to share, like, and remix outputs.
The quality of the initial outputs surprised observers who had been following AI music generation. Until Udio launched, Suno had been the obvious leader in vocal AI music with its v3 model. Reviewers comparing the two systems in mid-April 2024 broadly judged Udio's vocals and instrumental fidelity to be a step ahead, with cleaner highs, fewer audio artifacts, and more naturalistic phrasing on lead vocals. Rolling Stone described Udio as a direct rival to Suno that produced "shockingly good" output; The Verge called it "the AI music tool that has the music industry rattled"; and VentureBeat highlighted the breadth of styles the system could handle, from country to opera to drum and bass.
Music Business Worldwide reported that at peak usage during the first weeks of the public beta, Udio was producing roughly 10 songs per second across its user base. That throughput, paired with how closely some outputs resembled established recordings, primed the platform for the legal challenges that followed.
Udio's seed financing was announced alongside the April 10, 2024 launch. The headline number was $10 million in total seed funding, led by Andreessen Horowitz (a16z). Reporting from Music Business Worldwide and Digital Music News indicated that a portion of that round had closed earlier, in January 2024, before being topped up and disclosed publicly on launch day.
The cap table read like a deliberate combination of Silicon Valley credibility and music industry signaling. a16z brought the venture capital firepower and a track record of consumer AI bets. The musician investors brought name recognition and creative input, plus a defensive moat against the perception that Udio was purely a tech project hostile to artists. The Instagram and DeepMind veterans brought product and research advice.
| Investor | Type | Notes |
|---|---|---|
| Andreessen Horowitz (a16z) | Lead investor, venture capital | Marc Andreessen and Ben Horowitz have been vocal supporters of generative AI and consumer creator tools |
| UnitedMasters | Music distribution platform | Founded by Steve Stoute; distributes independent artists |
| will.i.am | Musician, producer, tech investor | Black Eyed Peas frontman and frequent backer of AI startups |
| Common | Musician, actor | Hip hop artist and Oscar-winning songwriter |
| Tay Keith | Music producer | Hit-maker behind tracks for Drake, Travis Scott, BlocBoy JB |
| Kevin Wall | Investor and live entertainment executive | Founded Control Room; produced Live Earth and Live 8 |
| Mike Krieger | Co-founder and former CTO of Instagram | Now Chief Product Officer at Anthropic |
| Oriol Vinyals | Head of Gemini at Google DeepMind | Personal investment, not a corporate Google check |
Udio has not publicly disclosed a Series A or any subsequent priced round through early 2026. Reporting on the UMG and WMG settlements made clear that the financial terms of those agreements were confidential, but the partnerships imply some form of equity or revenue participation by the labels in the new licensed platform. Compared with Suno, which raised a $125 million Series B in May 2024 and a $250 million Series C at a $2.45 billion valuation in November 2025, Udio has remained far more capital constrained on paper. The settlement strategy and the smaller capital base have led some industry observers to argue that Udio chose the licensing path partly because litigating against three majors simultaneously was not financially viable for a $10 million seed-stage company.
Udio has released several distinct model generations and feature drops since its launch. The internal naming has evolved from short tags tied to clip length (like Udio-32 and Udio-130) toward marketing-oriented version numbers (v1, v1.5, v1.5 Allegro). The table below captures the major milestones.
| Version | Release | Key changes |
|---|---|---|
| Udio v1 (Udio-32) | April 10, 2024 | Initial public beta; 32 to 33 second clips; text-to-music with vocals; extend forward and backward; built-in feed |
| udio-130 | June 2024 | New 2 minute generation model with clearer song structure (beginning, middle, end); audio upload feature for extending user audio; context window expanded from 30 seconds to 2 minutes |
| Audio Inpainting | May 13, 2024 | Select a portion of an existing track and regenerate it in place using surrounding context (28 second window); paid tier exclusive |
| Stems | 2024 | Separate finished outputs into vocals, bass, drums, and other instruments for export |
| Udio v1.5 | Late 2024 | 48 kHz stereo output; improved vocal warmth and phrasing; 2 minute context window; key control; audio-to-audio remixing; expanded language support; unified Creation page |
| Udio v1.5 Allegro | March 18, 2025 | Distilled version of v1.5; 4x faster generation with equal or better perceived quality; new default model; free for all users |
| Sessions | June 26, 2025 | Timeline editor with visual waveform sections (verse, chorus, bridge); DAW-style editing; Standard and Pro only |
| Walled-garden platform | Planned 2026 | Trained only on licensed catalogs from UMG, WMG, Merlin, Kobalt; downloads disabled; new subscription product |
The original public model produced clips of up to 33 seconds. Users wanting longer tracks built them by extending existing clips, either forward (adding a chorus after a verse) or backward (adding an intro before a verse). The 30 second context window meant that the model could only see the most recent 30 seconds of audio when generating the next section, which limited how much long-range musical coherence the system could maintain.
Despite those limits, the v1 model proved surprisingly capable across genres. Rolling Stone's launch review highlighted the model's ability to generate plausible barbershop quartets, baroque cantatas, dub reggae, trap, and 1970s soul from short prompts. The vocal model was strong enough that several listeners reported being unable to tell AI vocals from real performances on first listen.
In June 2024 Udio released what it called the udio-130 model, capable of generating up to 2 minutes of audio in a single pass. The release also introduced an audio upload feature, letting users seed generations with their own short recordings, and a fourfold expansion of the context window from 30 seconds to 2 minutes.
The practical upshot was that longer compositions began to sound more like songs. Where previous outputs sometimes drifted stylistically or harmonically over the course of a 4 minute extension, the new model could maintain key, tempo, and instrumentation across longer spans. DeepLearning.AI's The Batch newsletter covered the release, noting that the longer context made Udio more useful for actual song production rather than just short novelty clips.
Introduced on May 13, 2024 alongside the launch of subscription tiers, audio inpainting was Udio's first major editing tool and arguably one of the more technically novel features in consumer AI music. The system lets users select a portion of an existing generated track and regenerate just that segment based on the surrounding audio context.
In practice, users open a track, highlight a 28 second window around the section they want to change, then choose a narrower region inside that window for the AI to redo. Everything outside the selected region remains untouched. The technique is analogous to image inpainting in diffusion models, where masked pixels are regenerated conditioned on their unmasked neighbors. Common use cases include fixing a vocal line that sang the wrong word, smoothing an awkward transition between two extensions, or replacing an instrumental section that does not match the rest of the track.
Audio inpainting was launched as a Standard tier feature on desktop only and is generally credited with marking Udio as the more producer-oriented of the two major AI music platforms.
Released in late 2024, Udio v1.5 was the platform's first major model upgrade after the initial launch. The version represented improvements across multiple dimensions rather than a single new capability.
| Improvement area | Details |
|---|---|
| Audio quality | Output bumped to 48 kHz stereo with cleaner transients, better instrument separation, and improved coherence across long passages |
| Vocal realism | Warmer, less robotic vocal synthesis with better phrasing and breath modeling |
| Context window | Expanded from 30 seconds to 2 minutes for better long-form structure |
| Key control | Users can specify a musical key (e.g., C minor, A flat major) in the prompt; outputs largely respect the key request |
| Stems | Final outputs can be split into vocals, bass, drums, and other instruments |
| Audio-to-audio | Upload an existing recording and have Udio generate a new version in a different style |
| Language coverage | Expanded support for non-English lyrics including Japanese, Korean, Spanish, Hindi, and others |
| Unified Creation page | New interface combining prompt input, library, and editing tools on one screen |
Reviewers generally agreed that v1.5 closed the gap between AI-generated music and well-produced human recordings. Tom's Guide called the upgrade "massive" and said it was "the closest AI has come to making me think I was hearing real music." The increased context window made it easier to extend tracks to the full 15 minute ceiling without obvious stylistic drift.
Allegro launched on March 18, 2025, and Udio was careful to describe it not as a new model generation but as a distilled version of the existing v1.5 model. The Allegro release was the practical output of internal research on sampling step distillation, a technique that reduces the number of model evaluations needed to produce a single output without retraining the underlying weights.
The headline number was a 4x speedup at equivalent quality, with internal raters often preferring Allegro outputs to original v1.5 outputs on blind tests. Generation times for a 33 second clip dropped from roughly 30 to 40 seconds down to under 15 seconds, which meaningfully changed the feel of the product. Allegro became the new default for all users at no extra cost, while the original v1.5 model remained selectable as an alternative.
Introduced on June 26, 2025, Sessions is a major editing upgrade rather than a model change. It adds a timeline view to the Udio interface, with the waveform of a song displayed horizontally and individual sections (verse, chorus, bridge, intro, outro) auto-detected and color-coded. From the timeline view, users can extend a track from either end, replace a section by inpainting in place, swap takes from a stack of alternates, and adjust generation parameters per section.
The feature was widely interpreted as a move toward digital audio workstation (DAW) style production tools, which had previously been Suno's territory with Suno Studio. Sessions launched for Standard and Pro subscribers and remained behind the paywall as a clear premium differentiator.
The core capability of Udio is text-to-music synthesis. The user types a prompt describing the music they want and the system returns a complete audio file with both vocals and instrumentation. Prompts can be detailed ("upbeat 1970s funk in F minor with female lead, slap bass, wah-wah guitar, and synthesized brass stabs, about a robot in love with a vacuum cleaner") or short and stylistic ("baroque cantata in Latin"). The system handles a remarkably wide range of styles, including hip hop, country, classical, jazz, opera, gospel, drum and bass, lo-fi, and orchestral film score.
The system can run in three primary modes. The first is full song generation, where Udio produces both music and vocals based on a topic. The second is custom lyrics mode, where the user provides full lyrics and structure markers and the system generates music to fit. The third is instrumental mode, where vocals are omitted entirely. Style references can be supplied as artist names or genre descriptors, though Udio is selective about which artist names it permits in prompts following the copyright litigation.
Users build full-length songs by extending shorter generations. Forward extension adds new audio after the existing clip; backward extension adds new audio before it. Each extension respects the surrounding audio context (up to 2 minutes in v1.5) so that the new section continues the musical material rather than starting fresh. The total length ceiling for a single track grew over time from an initial 4 minutes to the current 15 minute limit, comfortably long enough for a progressive rock suite or an extended dance mix.
As described above, inpainting allows in-place regeneration of selected segments. The feature requires a paid subscription and is widely used by Udio power users to clean up vocal flubs, fix awkward transitions, or experiment with alternative ideas for a single bar or phrase without redoing the whole track.
Introduced with v1.5, audio-to-audio lets users upload an existing recording as a seed and ask the system to reimagine it in a different style. The output preserves the rough structure and melody of the input while shifting the timbre, instrumentation, or genre. Power users have used the feature to convert acoustic demos into produced tracks, or to generate alternative arrangements of their own material.
Udio can decompose a finished generation into four stems: vocals, bass, drums, and other instruments. Stems can be downloaded individually (when downloads are enabled) and used in external digital audio workstations for mixing, sampling, or live performance. The stems feature is one of the clearest indications that Udio's positioning skews toward producers and bedroom musicians rather than pure consumers.
With v1.5, users gained the ability to specify the musical key of a generation. The model does not always perfectly match the requested key, but the feature gives meaningful control over harmonic structure, particularly useful when generating a track to fit with an existing instrumental or vocal sample.
The Sessions feature, introduced in June 2025, adds visual editing on a timeline with auto-detected song sections. Users can rearrange parts, audition alternate takes, and stack edits without leaving the Udio interface. This positions Udio closer to a hybrid AI generator and lightweight DAW.
Udio automatically generates lyric videos with synced text overlays for finished songs, sized for sharing on TikTok, Instagram, and X. The lyric video function was launched in the first months after public release and remains a core social distribution feature.
Advanced users can specify the random seed used for a generation, which lets them reproduce a result exactly or systematically explore the latent space around a particular successful output. Seed control is a niche feature that matters mostly to power users running A/B experiments.
Udio offers three subscription tiers. Pricing and credit allocations reflect the values that took effect during the late 2025 licensing transition, when Udio doubled monthly credits on the Standard plan as a goodwill gesture to existing subscribers.
| Plan | Monthly | Annual (per month) | Monthly credits | Key features |
|---|---|---|---|---|
| Free | $0 | $0 | 10 daily plus 100 monthly bank | Basic generation, limited concurrent jobs, community access |
| Standard | $10 | $8 (billed as $96/year) | 2,400 | WAV downloads (when enabled), stem downloads, audio inpainting, Sessions, higher concurrency |
| Pro | $30 | $24 (billed as $288/year) | 6,000 | All Standard features plus bulk downloads (when enabled), highest priority queue, commercial use rights |
Credits do not roll over from month to month under standard plans. Add-on credit packs are available: 100 credits for $3 and 1,000 credits for $25, and unlike subscription credits these add-on credits do not expire. A 50% student discount is available on Standard and Pro for six months after subscribing, and Udio has periodically offered annual plan discounts in line with subscription seasonality.
During the 2025 to 2026 licensing transition, Udio temporarily disabled audio, video, and stem downloads across all subscription tiers. The downloads were expected to return once the new licensed platform launched in 2026, with the catch that downloaded content would only be usable inside the Udio ecosystem under terms that reflected the underlying license agreements with rights holders.
As partial compensation for the loss of downloads, Udio granted a one-time bonus of 1,000 non-expiring credits to all subscribers, raised the Standard plan monthly credits from 1,200 to 2,400, and raised the Pro plan monthly credits from 4,800 to 6,000.
On June 24, 2024, the Recording Industry Association of America (RIAA) announced that the three major record labels had filed two parallel copyright infringement lawsuits against the leading AI music generation companies. The case against Uncharted Labs, Inc. (doing business as Udio) was filed in the United States District Court for the Southern District of New York under the caption UMG Recordings, Inc. et al. v. Uncharted Labs, Inc., et al., docketed as case number 1:24-cv-04777 and assigned to Judge Alvin K. Hellerstein. The parallel case against Suno, Inc. was filed in the District of Massachusetts.
The plaintiffs in the Udio case included UMG Recordings, Inc., Sony Music Entertainment, and Warner Records, Inc., along with several subsidiary labels and affiliated entities. The complaint ran nearly 50 pages and was unusually detailed for an opening pleading, with extensive examples of allegedly infringing outputs juxtaposed against the copyrighted recordings they were claimed to have been copied from.
The core allegation was that Uncharted Labs had copied and ingested a vast corpus of copyrighted sound recordings to train its AI models without obtaining licenses or authorization from the rights holders. The complaint argued that this unauthorized copying constituted direct copyright infringement at multiple stages: first when the recordings were acquired and stored, then when they were processed for training, and again every time the resulting model generated an output that incorporated copyrightable elements of those original recordings.
A significant portion of the complaint focused on outputs. The plaintiffs provided numerous side-by-side examples where targeted prompts (referencing the decade, genre, lyrical themes, and stylistic markers of well-known recordings) caused Udio to produce outputs that closely resembled the named recordings. Examples in the public complaint included AI-generated versions that the labels argued mimicked recordings by Mariah Carey, the Temptations, ABBA, and many others. The complaint argued that these outputs were not just stylistically similar but exhibited specific musical fingerprints (riffs, chord progressions, vocal mannerisms, production quirks) that the model could only have learned by training on the originals.
The plaintiffs sought:
Udio filed its formal answer on August 1, 2024 (case 1:24-cv-04777, docket entry 26). The pleading admitted that the company had used copyrighted recordings in training but mounted a vigorous fair use defense rooted in the four-factor test under 17 U.S.C. section 107.
The central rhetorical move in the answer was an appeal to genre-level rather than recording-level use. "The premise of their case," Udio wrote, "is that musical styles, the characteristic sounds of opera, or jazz, or rap music, are somehow proprietary." The company argued that decades of judicial precedent established that no company controls a genre or style of music, and that the model learned general patterns and stylistic conventions rather than verbatim copies of any specific recording.
More broadly, Udio framed its training as a transformative use under fair use doctrine. The argument paralleled defenses raised in other generative AI lawsuits (notably the Authors Guild v. OpenAI cases and the New York Times v. OpenAI case): the company claimed that ingesting copyrighted works to extract statistical patterns for the purpose of producing new, distinct works qualifies as transformative in the same way that the search-engine indexing in Authors Guild v. Google was held to be transformative.
The RIAA pushed back hard. In a statement on the day Udio and Suno filed their answers, the trade group called the fair use argument a "major concession" because it confirmed that the AI companies had in fact trained on copyrighted recordings, an admission the labels had been working to extract through discovery. "There's nothing fair about stealing an artist's life's work, extracting its core value, and repackaging it to compete directly with the originals," the RIAA said.
The case proceeded through pleadings and early discovery throughout late 2024 and into 2025. Notable procedural events included motions to dismiss certain claims (denied in part), discovery disputes over the disclosure of Udio's training data, and protective orders covering financial and technical materials. Judge Hellerstein managed the docket and signaled in early hearings that he viewed the case as raising important questions of first impression about the application of fair use to large-scale generative AI training.
By mid-2025, both sides appeared to be preparing for a summary judgment motion that would test the fair use defense on its merits. The settlements that followed in October and November 2025 came before those motions were decided, meaning the fundamental question of whether AI music training on copyrighted recordings qualifies as fair use remains legally unresolved at the appellate level.
On October 29, 2025, Universal Music Group and Udio announced that they had reached a settlement of the copyright litigation and entered into strategic partnership agreements. The settlement was widely reported as a watershed moment for the AI music industry. The Rolling Stone headline framed it as Universal moving "from suing an AI company to partnering with it."
Key elements of the settlement, as disclosed in the joint press release and subsequent reporting, included:
Industry analysts described the deal as a hybrid that combined elements of a litigation settlement, a licensing agreement, and a strategic partnership. The Loyola Law School Entertainment Law Review's analysis argued that the deal "reframes fair use and AI in music," effectively setting a market price for AI training rights even as the underlying legal question of fair use remained unanswered. Manatt, Phelps & Phillips' client alert focused on what the deal would mean for artists, noting that the opt-in model could create a two-tier system of artists who participate (and earn from AI uses) and artists who do not (and miss out).
On November 19, 2025, Warner Music Group announced a separate settlement with Udio. The pattern closely mirrored the UMG deal. Both the underlying copyright claims and the path forward were resolved together.
Key elements of the WMG settlement included:
The Hollywood Reporter and TechCrunch both noted the speed with which Udio had moved from defendant to partner with two of the three majors. Within roughly three weeks, Udio had gone from facing potentially existential copyright liability to being positioned as the favored AI music partner for half of the global recorded music market.
Following the major label settlements, Udio extended its licensing footprint to indie labels and music publishers. Merlin, the global rights agency representing independent labels, announced a partnership with Udio in late 2025 covering its member labels. Kobalt Music Group, one of the largest independent music publishers, announced its own licensing agreement with Udio in April 2026. The Kobalt deal was particularly significant because it expanded Udio's licensing into the publishing side of music, covering compositions and songwriting rather than just recordings.
By May 2026, Udio's licensing roster included UMG, Warner, Merlin, and Kobalt, leaving Sony Music as the only major rights holder without a licensing agreement and still in active litigation.
As of mid-2026, Sony Music Entertainment remained the sole major record label still in active litigation against Udio. Reports from Digital Music News in December 2025 and again in early 2026 indicated that Sony Music had decided to continue litigating rather than accepting settlement terms similar to UMG's and Warner's, signaling either a higher valuation of its catalog rights or a strategic decision to push for legal precedent.
Discovery in the Sony Music case produced significant disclosures in the spring of 2026. In an April 2026 filing, Udio acknowledged that it had used audio scraped from YouTube to train its models, including the use of the open-source YT-DLP tool to download audio from the platform. The admission was significant because YouTube's terms of service forbid scraping, and the labels argued that the use of stream-ripping tools to acquire training data supported their claim that the underlying acquisition itself was unlawful regardless of any fair use defense at the model output stage.
The Sony v. Udio case is being closely watched as a likely vehicle for a definitive ruling on whether AI music training on copyrighted recordings qualifies as fair use. Both sides have signaled that they expect summary judgment motions and possibly appellate review before any final resolution, with substantive rulings projected for summer 2026.
Udio and Suno are the two dominant AI music generation platforms as of 2026. Their stories run remarkably parallel: both launched in roughly the same window, both raised serious venture capital, both attracted the same RIAA lawsuit on the same day, both have settled with at least one major label, and both have moved toward licensed catalogs and walled-garden distribution.
The two platforms differ in important ways, both technical and strategic. The table below captures the main distinctions as of early 2026.
| Feature | Udio | Suno |
|---|---|---|
| Founded | December 2023 | 2022 (incorporated); product launched 2023 |
| Headquarters | New York, NY | Cambridge, MA |
| Founders | 5 former Google DeepMind researchers | 4 former Kensho researchers (Mikey Shulman, Georg Kucsko, Martin Camacho, Keenan Freyberg) |
| Public launch | April 10, 2024 (closed beta from late 2023) | Late 2023 web app; earlier versions distributed via Discord |
| Funding | $10M seed led by a16z (April 2024) | $125M Series B (May 2024); $250M Series C at $2.45B valuation (Nov 2025) |
| Latest model | v1.5 Allegro (March 2025) | v5.5 (March 2026) |
| Audio quality | 48 kHz stereo; instrumental fidelity often rated highest | 44.1 kHz (v5+); emotionally expressive vocals; loudness and punch typically higher |
| Max song length | Up to 15 minutes | Up to 8 minutes (paid tiers) |
| Clip generation length | 33 seconds per clip; longer with udio-130 model | Varies by model; v5 generates longer single takes |
| Inpainting | Yes (audio inpainting, subscription only) | No direct equivalent at parity |
| Audio-to-audio | Yes (v1.5+) | Yes (Cover and Personas features) |
| Stem separation | Yes (4 stems: vocals, bass, drums, other) | Yes (Pro and Premier plans) |
| Key control | Yes (v1.5+) | Limited |
| Free tier credits | 10 per day plus 100 per month bank | 50 per day |
| Paid plans | $10/month, $30/month | $10/month, $30/month |
| Multitrack editor | Yes (Sessions, June 2025) | Yes (Suno Studio, Premier plan) |
| MIDI export | No | Yes (Premier plan) |
| Voice cloning | Restricted (artist licensing planned via UMG/WMG deals) | Yes (Voices in v5.5, March 2026) |
| RIAA lawsuit | Filed June 24, 2024 (SDNY, 1:24-cv-04777) | Filed June 24, 2024 (D. Mass.) |
| UMG settlement | Yes (October 29, 2025) | No (still litigating as of mid-2026) |
| WMG settlement | Yes (November 19, 2025) | Yes (November 25, 2025) |
| Sony status | Still litigating | Still litigating |
| Annual revenue | Not publicly disclosed | Reported around $200M (November 2025) |
| Valuation | Not publicly disclosed | $2.45B (November 2025) |
Reviewers tend to describe the two platforms' vocal models as having different sonic signatures rather than one being strictly better. Suno's vocals are often praised as more emotionally expressive, capturing breathiness, cracks, vibrato, and dynamic changes that sound closer to a real singer's performance. Udio's vocals are technically cleaner and more polished but can come across as a high-end MIDI vocal simulation, very controlled but lacking some of the imperfections that make human vocals feel alive. For pop, R&B, and singer-songwriter material, Suno is generally rated higher. For ambient, electronic, choral, and operatic material, Udio's precision works in its favor.
Udio's 48 kHz stereo output produces better instrument separation and tends to sound closer to a professional studio master. Reviewers consistently rate Udio higher for genres where instrumental fidelity and detail matter most: jazz, classical, film score, and dense electronic productions. Suno's instrumentation is good but tends to feel slightly compressed and radio-mastered, optimized for streaming loudness rather than audiophile detail.
Udio has positioned itself as the more producer-friendly platform with inpainting (since May 2024) and Sessions (June 2025) bringing DAW-style editing to the AI workflow. Suno's Studio (in the Premier plan) offers multitrack editing and MIDI export, addressing similar needs from a different angle. Each platform has its champions among different segments of the production community.
Suno generally outpaces Udio on raw generation speed, especially for longer outputs. Suno can complete 90 second songs in under 60 seconds, while Udio (even with Allegro) often needs 90 or more seconds for similar lengths. The gap narrowed significantly after Udio shipped Allegro in March 2025.
The two companies have made notably different strategic bets. Suno raised aggressively, expanded its consumer footprint, and litigated harder; Udio raised less, focused on producer-grade features, and was quicker to settle and pivot to a licensed model. Both approaches have defenders, and the eventual winner (if there is one) likely depends on how courts ultimately rule on fair use in AI training and how artists and consumers weigh licensed against unlicensed outputs.
Udio's most prominent moment in popular culture came in early May 2024, when comedian and content creator King Willonius used the platform to create "BBL Drizzy," an AI-generated R&B parody song mocking rapper Drake. King Willonius created the track on April 14, 2024, prompted by a Rick Ross social media post that joked that Drake looked like he had received a Brazilian Butt Lift. Willonius wrote comedic lyrics and prompted Udio to generate the song in a vintage 1970s soul style, with lush strings, a wandering bass, and a high tenor lead vocal.
The track went viral almost immediately, amassing more than 3.3 million streams on SoundCloud within a week. The phenomenon expanded well beyond the original track. Users across the internet began creating remixes and genre-swapped versions in styles including house, merengue, Bollywood, gospel, drum and bass, and classical, while instrumentalists posted covers on saxophone, guitar, harp, koto, and brass band. The track was particularly notable for showing how a single AI-generated piece could become a meme template that people across the world could remix.
Hip hop producer Metro Boomin further amplified the trend by remixing the track and launching the "BBL Drizzy" challenge on May 5, 2024. He released an instrumental titled "BBL DRIZZY 150 BPM.mp3" on SoundCloud, sampling Willonius's AI-generated vocal hook, and offered $10,000 and a free beat to the best lyricist who could build on the instrumental. Reports later suggested that Metro Boomin was initially unaware that the source vocal was AI-generated, which only deepened the conversation about authenticity and provenance in AI-era hip hop production.
The cultural significance of "BBL Drizzy" extended further when artists Sexyy Red and Drake collaborator Sexyy Red released "U My Everything" featuring Drake, which sampled the original AI-generated track. This is widely considered the first known commercially released song to sample an AI-generated piece of music, marking a milestone in the relationship between AI-generated content and the traditional music industry.
"BBL Drizzy" was also one of the specific outputs cited in the RIAA lawsuit's discussion of how Udio's outputs could resemble copyrighted recordings, and Spotify briefly removed the track from its platform amid the copyright debate. Willonius told Tulane's Hullabaloo and MusicTech in subsequent interviews that the experience showed both the creative potential of AI and the legal uncertainty surrounding it. "When I first got introduced to tools like ChatGPT, it unlocked all my creativity," he said. "But I also see why labels and artists are worried."
Other viral Udio moments included a wave of AI-generated diss tracks during the Kendrick Lamar and Drake feud in spring 2024, a steady stream of parody political songs in advance of the 2024 U.S. election, and a small but loyal community of producers using Udio for genre exercises that they would have struggled to produce in a traditional studio.
Udio launched with a built-in social feed that allowed users to publish, like, comment on, and remix tracks. The feed served both as a discovery surface and as a form of soft marketing, with viral songs surfacing on the home page and attracting new users.
The community spans a wide spectrum. On one end are professional musicians and producers using Udio for inspiration, demo work, or as a starting point for tracks they then take into a traditional DAW. On the other end are hobbyists, comedians, meme creators, and people with no musical training at all generating songs purely for entertainment. Between those poles sits a substantial middle of semi-serious creators who treat Udio as their primary creative outlet, releasing AI-generated music on streaming platforms (where allowed) and building followings on social media around their characteristic prompt styles.
The viral success of "BBL Drizzy" highlighted the platform's potential as a catalyst for participatory culture, where a single AI-generated track spawned thousands of remixes and covers across genres and instruments. That kind of distributed remixing is harder on Suno, which historically emphasized polished consumer outputs over remixable building blocks.
Udio also operates a public feedback portal where users submit and vote on feature requests, and the company has been responsive in announcing changelog updates roughly monthly. The community has been organized enough to lobby effectively for specific features, including longer track support and better key handling.
Udio has not published detailed technical papers describing its model architecture, but the platform's capabilities and the founders' DeepMind backgrounds give strong hints about the underlying approach. The platform almost certainly uses some combination of diffusion models and transformer backbones, with audio represented as latent codes (likely from a neural audio codec) rather than raw waveforms. This approach has been the dominant one in academic audio generation research since 2022 and is the same family of techniques used in MusicLM, AudioLM, and Lyria at DeepMind.
The system handles multiple challenging tasks simultaneously. It interprets natural language prompts and converts them into a conditioning signal. It generates instrumental arrangements that respect genre conventions across an enormous range of styles. It synthesizes vocals with realistic phrasing, breath, and prosody. It produces mixed and mastered stereo audio at 48 kHz, which is a meaningful technical achievement on its own. And it does all of this within a few seconds to a few minutes of generation time per output.
The context window improvements in udio-130 (June 2024) and v1.5 (late 2024) suggest that Udio's models have been getting steadily larger and more capable of long-range coherence. The Allegro release in March 2025 introduced sampling step distillation, which reduces the number of model evaluations needed per output by training a smaller student model to approximate the outputs of the full model in fewer steps.
The walled-garden architecture planned for the 2026 licensed platform implies additional technical infrastructure: content fingerprinting to identify and credit underlying source material, filtering to prevent generation of unlicensed copies, attribution metadata embedded in outputs, and presumably royalty accounting systems that track how often any given artist's voice or composition is invoked in outputs.
Udio's emergence, alongside Suno, has had outsized influence on debates about the future of music creation, copyright law, and the role of AI in creative industries.
By enabling anyone to produce professional-sounding music from a text description, Udio has lowered the barrier to music creation to near zero. People without musical training, instruments, or studio access can now generate songs that would have required tens of thousands of dollars in production budget five years ago. Supporters view this as a positive democratization of creative expression. Critics in the music industry argue it threatens the livelihoods of professional musicians, producers, and songwriters, and that the speed and scale of AI generation will flood streaming platforms with low-effort outputs that crowd out human-made music.
The RIAA lawsuits against Udio and Suno represented some of the most significant legal tests of how copyright law applies to AI training on creative works. The cases raised fundamental questions: does ingesting copyrighted recordings to train generative AI models constitute fair use; can the outputs of such models infringe on the copyrights of works used in training; what licensing frameworks might govern the use of copyrighted material in AI systems.
The settlements between Udio, UMG, and WMG have set an early market template even though they did not resolve the underlying legal question. The structure of those agreements (compensatory settlement, plus prospective license, plus revenue participation, plus opt-in artist compensation) is likely to become a reference point for similar deals across the AI industry. The Sony v. Udio case remains the only vehicle by which the courts might still produce a definitive fair use ruling for AI music training.
Udio's pivot from an unlicensed open-generation platform to a walled-garden licensed service reflects a broader industry trend. As legal and regulatory pressure mounts, AI music generators are moving toward partnerships with rights holders rather than relying purely on fair use defenses. This shift could reshape the competitive landscape, potentially favoring companies that secure the most comprehensive licensing agreements (which currently means Udio over Suno, at least on the recorded music side).
The model also raises questions about whether AI music generation will end up looking more like the music industry's existing licensing structure than like the open internet of the early 2010s. The walled-garden architecture, in particular, brings AI music closer to the model of platforms like TikTok and SoundCloud where rights are licensed at the platform level rather than at the user level.
Udio has faced several criticisms and challenges since its launch.
The central legal and ethical issue is the use of copyrighted recordings for training without permission. Udio's own admission in its August 2024 answer to the RIAA complaint that it had trained on copyrighted music defused some of the factual ambiguity but intensified the ethical debate. Even after the UMG and WMG settlements, the underlying questions remain contested: did Udio steal value from working musicians by training on their recordings without consent or payment; is the eventual opt-in compensation framework adequate to compensate the millions of artists whose works were used; what about artists who do not opt in or who are no longer alive to opt in.
Udio's vocal model is capable enough to produce convincing imitations of real singers' voices. The platform has implemented restrictions on certain artist names in prompts, but determined users have found ways around those filters. The UMG and WMG deals include explicit provisions on voice rights, but the broader cultural debate about whether AI voice cloning of human performers is acceptable (or whether it should be opt-in, opt-out, or banned entirely) remains unsettled.
While Udio's best outputs are striking, individual generations can vary widely in quality. Some outputs contain audio artifacts, off-key vocal moments, mispronounced lyrics, or arrangements that fail to follow the prompt. Power users routinely generate dozens of takes to find one keeper, which has practical implications for credit consumption and the value of the paid tiers.
The temporary removal of audio, video, and stem downloads during the 2025 to 2026 licensing transition has frustrated users who had relied on the platform for content creation workflows. Some power users have switched to Suno or other alternatives during the transition period, complicating Udio's positioning as the more producer-oriented platform.
While the UMG and WMG settlements include artist compensation frameworks, the actual payment rates have not been publicly disclosed. Some artists and industry advocates have expressed concern about whether compensation will be adequate, particularly for catalog artists whose voices or compositions might be heavily used by the AI but who receive only a small share of any resulting revenue.
Suno's significantly larger funding, higher valuation, and larger user base present real competitive challenges. With roughly $10 million in disclosed seed funding against Suno's nearly $400 million, Udio has had to be far more capital efficient and has placed bigger strategic bets (notably the licensing pivot) to compensate. The settlement-driven equity participation by UMG and WMG may help, but the company's runway and competitive position remain a subject of industry speculation.
The spring 2026 disclosure that Udio had used YT-DLP to scrape audio from YouTube added another layer of concern. Even if the eventual outputs are protected by fair use, the methods used to acquire training data implicate YouTube's terms of service and potentially additional legal exposure separate from the music copyright claims.
| Date | Event |
|---|---|
| December 2023 | Uncharted Labs (Udio) founded by five former Google DeepMind researchers |
| January 23, 2024 | Initial portion of seed funding closed with Andreessen Horowitz |
| April 10, 2024 | Public beta launch at udio.com; $10M total seed round announced |
| April 14, 2024 | King Willonius creates the original "BBL Drizzy" track on Udio |
| May 5, 2024 | Metro Boomin releases his "BBL DRIZZY 150 BPM" instrumental and launches the remix challenge |
| May 13, 2024 | Audio inpainting feature launched alongside subscription tiers |
| June 2024 | udio-130 model released; audio upload feature; 2 minute context window |
| June 24, 2024 | RIAA files copyright lawsuit (UMG, Sony, Warner v. Uncharted Labs) in SDNY, case 1:24-cv-04777 |
| August 1, 2024 | Udio files answer asserting fair use defense (docket entry 26) |
| Late 2024 | Udio v1.5 released with 48 kHz output, key control, audio-to-audio, expanded language support |
| March 18, 2025 | Udio v1.5 Allegro launched: 4x faster generation, new default model |
| June 26, 2025 | Sessions timeline editor launched for Standard and Pro subscribers |
| October 29, 2025 | Settlement with Universal Music Group announced; strategic partnership for licensed platform |
| November 19, 2025 | Settlement with Warner Music Group announced; licensing agreement signed |
| November 25, 2025 | Several plaintiffs in the original complaint voluntarily dismiss claims following settlements |
| Late 2025 | Audio, video, and stem downloads temporarily disabled across all plans during transition |
| Early 2026 | Merlin licensing agreement announced for independent labels |
| April 2026 | Kobalt Music Group licensing agreement announced, expanding into publishing |
| April 2026 | In Sony Music litigation, Udio acknowledges using YT-DLP to scrape audio from YouTube for training |
| 2026 (planned) | Launch of new fully licensed AI music creation platform |