Suno

Suno was founded in 2022 in Cambridge, Massachusetts by four colleagues from Kensho, a financial-AI company, where they had worked on transcribing earnings calls and came away convinced that audio was technically richer than text and badly underserved. Its public roots are open source: before the consumer app it released Bark, a transformerTransformerThe neural network architecture built on attention that underpins modern language and, increasingly, image and video models. It scales well and handles long-range relationships better than convolutional networks. text-to-audio model, in 2023. The web app arrived in December 2023, and the funding tells the trajectory, a 250-million-dollar round at a 2.45-billion valuation in November 2025 followed by over 400 million at a 5.4-billion valuation in June 2026, with reporting of roughly 300 million dollars in annual revenue and users generating over 7 million tracks a day.

The model line moved fast: v3 in early 2024 brought it mainstream, v4 added Personas to reuse a track's vocal character, v4.5 stretched songs to eight minutes, and v5 and the current v5.5 improved fidelity and added Custom Models, voice cloningVoice cloningReproducing a specific person's voice from a short sample so it can speak new lines, often in any language. through a feature called Voices, and a full generative workstation called Suno Studio with stemsStemsThe separated individual tracks of a piece of music, such as vocals, drums, and bass, which can be edited or remixed on their own. and MIDI export. On architecture, the honest read matters: the current commercial models are undisclosed, and the widely repeated idea that Suno runs on separate Bark and Chirp engines describes the early open models, not v5. What is confirmed, from the founders themselves, is that Suno treats music as a tokenTokenA single unit of input, a word fragment for text or a patch for images, that a model processes.-prediction problem, generating vocals and instrumentation together as audio rather than composing symbolic notes and rendering them, which is why its vocals are the standout and also why precise, DAW-style control remains hard.

Suno lives on the web and in mobile apps, with a free non-commercial tier and paid plans that grant commercial rights and the advanced tools. It is best at catchy, complete songs with vocals, quick demos, genre pastiche, and jingles, and weakest at reproducing an exact arrangement, flawless pronunciation of unusual words, and anything requiring guaranteed copyright cleanliness. The prompting craft is specific: put genre, mood, tempo, instrumentation, and vocal type together in the style box, put your actual lyrics in the lyrics box rather than the style box, use structure tags like verse and chorus, then generate several times, pick the best seed, and refine with Extend, Covers, Personas, or Studio rather than expecting one shot to be perfect.

On June 24, 2024 the three major labels, coordinated by the RIAA, sued Suno in Massachusetts, alleging it copied enormous quantities of copyrighted recordings to train its model. Suno's defense is that training is transformative fair useFair useThe US legal doctrine that permits some unlicensed use of copyrighted work, and the central defense labs raise for training models on scraped data. Its application to AI training is still unresolved., and a September 2025 amended complaint added claims that it circumvented YouTube's protections to download audio, pushing the theoretical damages into the billions. Then the posture flipped. Warner settled in late 2025, in a deal that notably included Suno acquiring Warner's Songkick ticketing platform, and Suno committed to launch licensed models in 2026 and retire the old ones, with artists able to control use of their name and voice.