What it is
Suno is a website and phone app that makes complete songs from a short written description. You type something like "an upbeat folk song about a dog who hates bath time", press a button, and within a minute or so you get two versions of a finished track, with a singer, lyrics, instruments, a verse and a chorus, all generated by AI. You can write the lyrics yourself or let Suno write them, and you can describe the style in as much detail as you like: the genre, the mood, the tempo, the kind of voice.
The company was founded in 2022 in Cambridge, Massachusetts, by four people who had worked together at Kensho, an AI company owned by S&P Global: Mikey Shulman, who became chief executive, Georg Kucsko, Martin Camacho and Keenan Freyberg. All four were amateur musicians. Their first public release, in April 2023, was not a music app at all. It was Bark, a free, open-source model (one whose code anyone can download and use) that turned text into audio: speech in many languages, laughter and sighs, simple sound effects and snatches of music. Bark showed the team could make believable sound from text. Suno then turned that ability towards songs.
Suno reached a much wider audience in December 2023, when Microsoft added it to its Copilot chatbot, so anyone could ask Copilot for a song and get one back. Suno's own site improved quickly after that. According to its release notes, version 3 (spring 2024) could make songs of about two minutes, version 3.5 (May 2024) reached four minutes, version 4 arrived in November 2024, version 4.5 in May 2025 and version 5 in September 2025. In September 2025 the company also launched Suno Studio, a music-editing workspace where you can generate and rearrange individual parts, such as just the drums or just the vocal (these separate parts are called stems).
How does it work? Suno has not published the details of its song models. The general idea is the same as other generative AI: a system is trained on a very large collection of recordings until it learns the patterns of music, such as how a chorus lifts, how a bass line follows the chords, how a singer's voice shapes a word. It then produces new audio that follows those patterns, guided by your description. It does not stitch together clips from existing songs, but, as the court cases below show, exactly what it learned from, and whether it can reproduce what it learned, is at the heart of the arguments about it.
Product details change fast, so treat everything in this paragraph as true at the time of writing and check Suno's site yourself. In September 2026 Suno replaced all its older models with a new "v6" family. Paying subscribers get v6 and a less predictable variant called v6-wild. Free users get a faster, lighter model called v6-mini. There are three plans: a free plan (Basic) and two paid plans, Pro and Premier.
Why it matters
It puts songwriting within reach of people who can't play an instrument. Before tools like Suno, getting a finished, sung song meant learning an instrument, hiring musicians, or buying from a library of stock music. Now a teacher can make a song to help a class remember the planets, a family can make a birthday song with the grandchild's name in it, and a small business can try out a jingle idea in an afternoon. In November 2025 Suno said nearly 100 million people had made music on it in two years, and TechCrunch reported the company was earning about 200 million dollars a year, mostly from subscriptions. Those are the company's own figures, but they show how many people want this.
It is at the centre of the fight over what AI may learn from. Suno trained its models on recorded music without asking the owners first, and said openly that it had done so. The world's three biggest record companies sued it. Collecting societies (the organisations that gather royalties for songwriters) in Germany and Denmark sued it too, as did independent musicians. How these cases end will shape whether AI music companies must pay for what they learn from, much as the cases over pictures and books are doing for images and text.
It is changing what reaches your ears. AI-made songs are now uploaded to streaming services in huge numbers. Deezer, a French streaming service, said in April 2026 that it was receiving nearly 75,000 fully AI-generated tracks a day, about 44% of all new uploads, and named Suno and Udio as the most common sources. Deezer also said those tracks made up only 1 to 3% of listening, and that it had flagged around 85% of the streams they did get as fraud: fake plays designed to collect royalties. Streaming services have responded with labels and filters, described below.
It raises the question of who owns an AI song. If you want to sell your song, put it on Spotify, or use it in a paid advert, you need to know what you're allowed to do. The answer depends on which plan you used and on copyright law, which in many countries protects only what a human made. That is covered in "Common mistakes" below.
Real-world examples
The record labels sue, and Suno admits what it trained on (2024 to 2026)
On 24 June 2024 the Recording Industry Association of America, the trade body for the US record business, announced two lawsuits on behalf of the three major label groups, Universal Music Group, Sony Music and Warner Music Group. One was against Suno, in federal court in Massachusetts. The other was against its rival Udio, in New York. Both accused the companies of copying huge numbers of recordings without permission to train their AI, and asked for damages of up to 150,000 dollars for each song infringed.
At first Suno would not say what its training data was. In its formal reply to the court on 1 August 2024, it changed course. The filing said the tens of millions of recordings its model was trained on "presumably included" recordings owned by the labels. Suno argued this was fair use, the part of US law that allows some copying without permission, for example for criticism or to make something new. Mikey Shulman wrote in a blog post the same day that Suno trained on music it could find on "the open internet", and compared what the AI does to a child learning rock music by listening to it. The RIAA replied that this was a major concession of facts Suno had tried to hide.
The case then turned to how the music was collected. In September 2025, Universal and Sony (by then without Warner, see below) amended their complaint to accuse Suno of "stream ripping": downloading audio from YouTube by getting around YouTube's protections against downloading. On 18 August 2026 the judge, F. Dennis Saylor IV, allowed that claim to go ahead. The case had not reached a verdict at the time of writing. A separate proposed class action on behalf of independent artists, filed in June 2025 by country singer Tony Justice, also survived Suno's attempt to have most of it thrown out in August 2026.
Outside the US, Suno lost its first court judgment. On 31 July 2026 a court in Munich ruled in favour of GEMA, the German collecting society for songwriters, finding that Suno's models had memorised six songs in GEMA's catalogue, including "Rasputin", "Daddy Cool" and "Mambo No. 5", and could reproduce them. Suno said the ruling mischaracterised how its technology works and that it was considering an appeal. Denmark's collecting society, Koda, filed a similar case in Copenhagen in November 2025.
Warner changes sides, and Suno rebuilds its models (2025 to 2026)
On 25 November 2025, Warner Music Group settled its lawsuit against Suno and became its partner instead. Under the deal, Suno agreed to launch new models trained on licensed music in 2026 and to retire its existing ones, to let Warner artists and songwriters choose whether their voices, names and songs could be used, and to limit downloads so that only paying users could take songs off the platform. Suno also bought Songkick, Warner's concert-listing service. The week before, Suno had announced it had raised 250 million dollars in new investment, valuing the company at 2.45 billion dollars. Udio had settled with Universal and with Warner in the autumn of 2025; that story is on the Udio page.
The promised models arrived on 9 September 2026 as the v6 family. Suno said they were trained from scratch on a new set of data including music licensed from Warner, from the music company BMG, and from artists who distribute through Believe and its service TuneCore. Warner's chief executive, Robert Kyncl, said its artists and songwriters would now share in the revenue from Suno's paying subscribers. Every older Suno model was retired. Download limits began a week earlier, on 3 September: 20 songs a month on Pro, 60 on Premier, and a handful of one-off trial downloads for existing free users. Reaction from users was mixed, with some welcoming the new sound and others complaining about the limits. Universal and Sony had not signed and were still suing.
This example matters because it shows one possible way out of the copyright fight: companies that trained without permission paying to license music after the fact. Critics, including some musicians and lawyers, argue that this lets a company profit first and settle later. Supporters say it is the fastest way to get artists paid. Both views were being argued at the time of writing.
An AI singer on the Billboard charts (2025)
In September 2025, a singer called Xania Monet appeared on Billboard's charts for the first time, with the gospel song "Let Go, Let God" and the R&B song "How Was I Supposed to Know", which reached number one on Billboard's R&B Digital Song Sales chart. Xania Monet is not a real person. She was created by Telisha Jones, a poet and designer from Mississippi, who writes the lyrics herself and uses Suno to turn them into sung, produced tracks, with an AI-made image as the singer's face. According to Billboard, bidding for her reached 3 million dollars, and she signed with Hallwood Media, a company led by the former Interscope executive Neil Jacobson. Some well-known R&B artists, including Kehlani, criticised the idea of an AI singer taking chart places and deals from human ones.
The case shows both sides of the tool. A person with real lyrics and no musical training reached an audience she could not have reached before. It also shows why streaming services began labelling AI artists: in August 2026 Spotify announced an "AI Persona" badge for artist profiles that represent AI-generated identities rather than real people.
An everyday scenario: a jingle for an author's podcast
This is an illustration, not a real case. Priya writes cosy mysteries and has started a short weekly podcast about writing. She wants ten seconds of music to open each episode and a longer theme for her next book trailer. She writes four lines of lyrics herself, including the podcast's name, and asks Suno for "warm acoustic jazz, brushed drums, upright bass, female vocal, playful, 90 beats per minute". Of the first two results, one is lovely but the singer swallows the podcast's name, and the other rhymes neatly but sounds too much like a coffee advert. She regenerates, changes "playful" to "wry", and gets a version she likes. She is on the Pro plan, so she downloads it with commercial rights, trims the intro in a free audio editor and uses the full song under her Runway-made trailer. Because she wrote the lyrics, those words are hers. The music may not be protected by copyright, so she knows someone else could, in theory, reuse it.
Try it yourself
Buttons and menus change often, so use these steps as a guide to the process. At the time of writing Suno has a free plan that gives 50 credits a day, which its help pages say is enough for about 10 songs. Free songs can be played and shared on Suno but, since September 2026, generally not downloaded, and they are for personal, non-commercial use only.
- Sign up for a free account at suno.com. You need to sign in before you can create songs.
- Start with a simple description. Click Create and type one sentence describing a song, such as:
A cheerful 1960s-style pop song about finally finishing the first draft of a novel, handclaps, bright horns, male and female duet - Listen to both versions. Suno normally gives you two. Notice what changed between them: the melody, the voice, how the lyrics were fitted to the music.
- Switch to writing your own lyrics. Look for the custom or advanced mode. Paste in a short verse and chorus you have written, and put the style in its own box. Marking sections helps, for example: “
[Verse] Coffee's cold and the cursor's blinking Chapter twelve and I'm overthinking [Chorus] One more page, one more page“ - Describe sound, not stars. Suno does not accept artist names in style prompts and removes them. Describe what you hear instead: genre, era, instruments, tempo, mood, type of voice. "Smoky female alto, slow blues, Hammond organ" works better than a famous name.
- Refine one thing at a time. If a word is sung wrongly, change its spelling or split it with a hyphen. If the song rushes, ask for a slower tempo. If the chorus is weak, rewrite the chorus lyrics rather than the whole prompt. This step-by-step approach is iterative refinement.
- Check the rules before you use it publicly. On the free plan, keep it personal. If you plan to use a song in anything that earns money, including a monetised YouTube channel, a paid course or a book advert, read Suno's current terms and expect to need a paid plan when the song is made.
Common mistakes and misunderstandings
Assuming the free plan lets you use songs commercially. At the time of writing, Suno's help pages say that on the free plan Suno owns the songs and you may use them only for non-commercial purposes. Songs made while on Pro or Premier belong to you and come with commercial use rights, which you keep even if you cancel. Upgrading later does not give commercial rights to songs you made on the free plan. Terms have changed several times, including in September 2026, so read the current version yourself.
Thinking "I own it" means "it's protected by copyright". These are different things. Suno's terms decide who owns a song between you and Suno. Copyright law decides whether anyone can stop others copying it. In the United States, material made entirely by AI from a prompt is not protected, which is the same position the US Copyright Office has taken on images (see Midjourney). Suno's own help pages say a song made 100% with AI would not qualify, but lyrics you wrote yourself are yours. Other countries may treat this differently.
Believing an AI song is automatically "original". Suno says it blocks artist names and does not aim to copy existing songs, but the Munich court found its older models could reproduce specific songs. If something you make sounds very close to a song you know, don't release it. A soundalike can be taken down and can land you in a copyright dispute.
Expecting to upload unlimited songs to streaming services. Streaming services are pushing back against mass uploads of AI music. Spotify said in September 2025 that it had removed more than 75 million "spammy" tracks in a year, launched a spam filter, and only allows AI voice impersonation of an artist with that artist's permission. Deezer tags albums containing fully AI-generated tracks and keeps them out of its recommendations. Distributors (the services that put your music on Spotify and Apple Music) have their own rules too. Check them before you plan a release strategy.
Thinking you can make a song sound like a real singer. Imitating a real person's voice without permission breaks Suno's rules and Spotify's, and can break the law in some places. The deals with Warner and others are built around artists choosing to take part, not around anyone copying anyone.
Treating the first result as finished. The first version often has mumbled words, lyrics that don't scan, an ending that fades oddly or stops mid-note, or a chorus that never quite arrives. Good results usually take several attempts and some editing, and a human ear is still the only judge of whether it's good.
Forgetting the argument around it. Many musicians object to tools trained on their work without permission, and some listeners avoid AI music altogether. If you use Suno in public work, being open about it, for example in your podcast notes or video description, is honest and increasingly expected.
Related elements
- Udio: Suno's closest rival, sued by the same labels on the same day, with its own settlements.
- ElevenLabs: AI voice and speech, often used alongside AI music for narration and dubbing.
- Midjourney: where the US Copyright Office's rulings on AI-made work are explained in detail.
- Runway: AI video, for the pictures that go with an AI-made soundtrack or book trailer.
- Datasets: the collections of training material at the heart of Suno's lawsuits.
- Prompt: the description that tells Suno what kind of song to make.
- Iterative Refinement: improving a song one change at a time instead of starting over.
- Writing Assistance: AI help with the lyrics before they are turned into a song.
- Canva AI: simple design tools for the cover art and videos that often accompany a song.