Suno

Suno develops AI models and tools for making music, from turning a simple idea into a song to editing, sampling and arranging audio. Mikey Shulman is its co-founder and CEO. On September 9, 2026, the company introduced v6, a family of music models developed with partners including Warner Music Group, BMG and Believe.[3][10]

The v6 release brought plain-language editing of individual song sections, mashups from multiple sources, sampling and instrument isolation in one workflow, and creation from text, audio, images and video. It also supports changing a single lyric without rebuilding the whole song. The flagship v6 and the more exploratory v6-wild serve Pro and Premier subscribers; v6-mini is available to everyone. Suno announced that it would retire earlier models as v6 rolled out.[10]

Suno’s earlier speech and generative-audio work provides context for its music products: before building the song-making service, the founders worked together on text and speech recognition, then released Bark, a model capable of generating both speech and music.[1][2]

Shulman and his three co-founders—Georg, Martin and Keenan—previously worked together at Kensho. Their work was largely text-based until a financially focused speech-recognition project led them into audio. They subsequently left Kensho to pursue generative audio full time. In a March 2024 interview, Shulman recalled that others encouraged them to focus on speech, but the founders, all musicians and music enthusiasts, kept experimenting with music models because they enjoyed making music.[1]

Suno released Bark on April 20, 2023, and announced its MIT license on May 1, allowing commercial use. Bark is a transformer-based text-to-audio model that generates multilingual speech, music, background noise, sound effects and nonverbal sounds such as laughter and sighs. Its generative design also means that output can depart from the supplied script.[2]

Bark uses a GPT-style architecture and quantized audio representations from EnCodec. It converts text directly into audio without an intermediate phoneme representation, allowing prompts to extend beyond spoken words. Its repository acknowledges nanoGPT, EnCodec, AudioLM and Vall-E; Shulman specifically credited Andrej Karpathy’s nanoGPT for borrowed code. He described Suno’s contribution as turning audio into tokens so that transformer training could learn from broad audio rather than relying solely on narrowly paired text-to-speech data.[2][1]

Bark already allowed users to encourage singing by placing music-note symbols around lyrics. Shulman recalled seeing people in Suno’s Discord trying to extract music from the speech model, which he interpreted as evidence of an appetite for music creation.[2][1]

Suno also collaborated with NVIDIA on speech recognition. Parakeet RNNT 1.1B, jointly developed by the NVIDIA NeMo and Suno teams, is an English FastConformer Transducer model with approximately 1.1 billion parameters, licensed under CC-BY-4.0. Its model card reports training on 64,000 hours of English speech prepared by both teams: 40,000 private hours and 24,000 hours from public datasets. Shulman discussed the collaboration in March 2024.[12][1]

Suno gained attention with a new website in December 2023. In May 2024, Shulman described its first song-making product as having launched eight months earlier, a separate milestone from Bark’s release. By March 2024, the service offered a single-box description mode and an expert mode for supplying and adjusting lyrics and style. Shulman reported that expert mode accounted for more than half of usage.[1][3][2]

Shulman described two uses: quickly making a song about a moment, and repeatedly refining a song until it matched an idea in the creator’s head. His example of the first was a coffee order for co-founder Martin that came back labeled “Margoo”; he recalled making a song about it in five seconds that stayed in the team’s heads. Suno’s March 2024 focus was broad consumer participation rather than professional production workflows, although Shulman said professionals were already using it for inspiration and samples.[1]

In that interview, Shulman explained Suno’s modeling approach as learned audio tokenization followed by transformer next-token prediction: “Let me predict the next tiny little bit of audio”. The team sought to minimize imposed musical structure and let models learn it. Streaming imposed a practical constraint on model size: generation had to produce audio at least as fast as listeners could hear it. He distinguished three evaluation axes—sound fidelity, song quality and controllability—and emphasized listening alongside quantitative benchmarks through the phrase “aesthetics matter”.[1]

V3 Alpha reached Pro and Premier subscribers on February 22, 2024, with a two-minute maximum clip length; v3.5 became available to everyone on May 30 with four-minute songs. Suno launched its iOS app on July 1, initially in the United States, and its first Android version on December 5. Between those mobile releases, v4 launched in beta for Pro and Premier on November 19. It introduced Remaster for older tracks and ReMi lyrics assistance, and powered Covers for reinterpreting uploaded audio and Personas for carrying a track’s vocals, style and vibe into subsequent creations.[11][7]

V4.5, released May 1, 2025, extended song generation to eight minutes, added a helper that expands genre ideas into detailed style descriptions, and allowed Covers and Personas to work together. V5 followed on September 23. Two days later, Suno launched Studio for Premier subscribers, adding generated stems, a multitrack timeline, controls for tempo, volume and pitch, and audio or MIDI export for continued work in another digital audio workstation.[8][11]

V5.5, released March 26, 2026, introduced Voices, Custom Models and My Taste. Voices lets creators use their own voices; its announced verification process matches singing in uploaded or recorded audio against a randomly requested spoken phrase. At launch, Voices were private to their owners and available to Pro and Premier subscribers, who could also create up to three custom models tuned to their own music. My Taste, available to everyone, learns genre and mood preferences. In August, Voices reached iOS and Android with trials on free plans.[9][11]

The professional audience became an explicit part of Suno’s plans: its November 2025 financing announcement promised more sophisticated professional tools alongside casual creation, and v5.5 was designed to include working professionals. Studio 2.0, released August 13, 2026 for Premier subscribers, expanded the browser-based workstation with MIDI recording and editing, audio effects, a wavetable synthesizer, automation and a chat bar for generating instruments and vocals. A September 17 update added further MIDI capabilities, including creating audio stems from MIDI clips through chat.[4][9][11]

On November 25, 2025, Suno and Warner Music Group announced a partnership to develop licensed AI music models. The agreement settled previous litigation between the two companies. Suno also acquired Songkick, the live-music and concert-discovery platform, from WMG and committed to continuing its operation.[6]

The partnership committed to artist and songwriter control over whether and how their names, images, likenesses, voices and compositions would be used in new AI-generated music. It also announced future platform changes: retirement of existing models when new licensed models launched, no downloads for future free-tier songs, and monthly download caps for paid users with additional downloads available for purchase. These were announced policies; the September 2026 v6 release still described model retirement as part of the rollout. That release also described artist-specific opt-in experiences, with payment for participating artists, as products under development.[6][10]

On May 21, 2024, Suno announced that it had raised $125 million, naming Lightspeed Venture Partners, Nat Friedman and Daniel Gross, Matrix and Founder Collective as partners. It said the money would accelerate product development and team growth. At that date, the company reported that 10 million people had made music using Suno.[3]

On November 19, 2025, Suno announced a $250 million Series C at a $2.45 billion post-money valuation. Menlo Ventures led, with participation from NVentures, NVIDIA’s venture capital arm, Hallwood Media, Lightspeed and Matrix. Suno reported that nearly 100 million people had made music on the platform over the preceding two years—a cumulative participation figure rather than an active-user or subscriber count.[4]

On June 3, 2026, Suno said it had raised over $400 million in Series D funding at a $5.4 billion post-money valuation. Bond Capital led the round, alongside IVP, Forerunner, Union Square Ventures, Alkeon and Quiet, with participation from existing investors Matrix, Lightspeed, Menlo Ventures and Schroders Capital.[5]