AI songs are going viral. Make yours.
Lyro is an AI music studio. Describe a vibe, or bring your own vocal or beat, and it writes, sings and produces the song. Then you mix and master it in the same place, ready for TikTok and streaming.
Tap the squares, then press play. This is the rhythm under half the songs on your feed.
Turn this groove into a songHear itMade on Lyro
Raw model output, cut to forty seconds.
Every clip came out of the same pipeline you get. Open the prompt behind any of them and make your own version.
Deep house, 120 BPM, soft rounded kick, shuffled hi-hats with swing, warm analog sub bass, lush Rhodes chords with jazzy seventh voicings. Soulful female vocals sung in Turkish, warm and emotive. Model: Lyria 3 Pro
Afro house, 122 BPM, four-on-the-floor kick with deep rolling log drum bassline, congas, shakers, djembe, warm marimba motif, airy pads. Female vocals sung in Turkish, soft breathy delivery. Model: Lyria 3 Pro
Afro house, 122 BPM, log drum bassline, layered organic percussion, kalimba motif. Short chopped and pitched vocal phrases used rhythmically as a hook, heavy reverb and delay. Model: ACE-Step 1.5
Melodic organic house, 121 BPM, driving soft kick, plucky arpeggiated synth motif, evolving analog pads, hand drums, emotional minor-key progression. Male vocals sung in Turkish, deep calm delivery. Model: Lyria 3 Pro
Afro house, 122 BPM, rolling log drum bassline, congas, shakers, djembe, rim clicks, warm marimba and kalimba motif, airy atmospheric pads, spacious modern club mix. Instrumental only. Model: Lyria 3 Pro
Why nowChecked 21 Sep 2026
Half of all new music is already made this way.
of the new tracks delivered to Deezer each day were fully AI-generated at the June 2026 peak: about 90,000 a day.
Deezer, 21 Jul 2026of 9,000 listeners in eight countries could not tell a fully AI-generated track from a human one in a blind test.
Deezer and Ipsos, 12 Nov 2025The Velvet Sundown passes 1.4 million monthly listeners on Spotify, then confirms it is “composed, voiced, and visualized with the support of artificial intelligence”.
Berklee, 22 Jul 2025In a blind test of 9,000 people across eight countries, 97% cannot tell AI-generated music from human-made music.
Deezer and Ipsos, 12 Nov 2025Suno reports 2 million paying subscribers and $300 million in annual recurring revenue.
TechCrunch, 27 Feb 2026Spotify and Universal Music Group sign licensing deals for a tool that lets fans make AI covers and remixes.
Spotify, 21 May 2026“Papaoutai (Afro Soul)” reaches 210.7 million streams in six months, the highest-ranking AI-assisted track in the world. It is an afro record.
Luminate Midyear Report 2026Fully AI-generated music passes half of everything delivered to Deezer: around 90,000 tracks a day.
Deezer, 21 Jul 2026
The flood is mostly forgettable. On Deezer those uploads earn 1 to 3% of streams, and Luminate notes that no AI track has yet changed how people listen. Making music is solved. Making something people replay is not. That gap is the opening.
Start from anythingFour ways in
Bring nothing, a vocal, a beat, or both.
- Describe itA sentence is enough. The prompt generator turns it into a producer’s brief: tempo, drums, bass, chords, vocal.
- Write the words, or skip themBring lyrics, let the writer draft them in your language, or go instrumental.
- Generate, then polishLyro sends the job to the model that suits it: one for sung lyrics, another for chops and tempo-locked beats. Finish it in the mixer.
- Upload or record itWAV, MP3, M4A, or straight from your microphone.
- Lyro reads the tempo and keyDetected in your browser in a few seconds. Correct it if you know better.
- A beat is written around itThe instruments are composed to your vocal, so they follow your phrasing instead of fighting it.
- Upload the instrumentalLyro finds its tempo and key so everything after lines up.
- Write the song, or let it improviseWith lyrics, Lyro sings them at your beat’s tempo and key and hands back the isolated vocal. Without, it improvises a topline over your track.
- Mix them togetherBoth parts open in the mixer, lined up on bar one.
- Drop in your tracksVocal, beat, extra stems. Waveforms snap to the beat grid.
- Balance, shape, spaceGain, pan, EQ, compression, reverb, delay, and a tempo-synced pump for that house breathing.
- Master to a targetPick streaming or club loudness, check the A/B, export WAV.
The studioAfter the first draft
Everything after the first draft, in one place.
Prompt generator
Pick a genre, a mood and a few instruments, and get a brief a model actually follows. Try the free version
Lyric writer
Drafts hooks, verses and vocal-chop phrases in twelve languages, then tightens or translates what you already have.
[Hook]
Stay a little longer
The night is not done
Stem splitter
Pull the vocals, drums, bass and everything else out of any track, yours or generated.
Vocal tools
Record in the browser, clean up with presets, shift the pitch, stretch to tempo.
Mixer
A multitrack timeline that snaps to the beat. Every track gets EQ, compression, sends and pump.
Mastering
Loudness targets for streaming and club, a limiter, and an A/B switch so you hear the difference.
-14.0 LUFS integrated
Share page
An unlisted link with artwork and a player. Send it to a friend or put it in your bio.
Tested, not guessedOur blind test
We listened blind before we chose which model sings your song.
Lyro picks between models by what the track needs, above all whether the words have to be understood. The choice comes from a listening test where the model names were hidden. It was a small one, and we publish how small.
| Model | Pronunciation | Production | Overall | What we do with it |
|---|---|---|---|---|
| Lyria 3 Pro | 4.67 | 4.43 | 4.57 | Routed by default for sung lyrics |
| ACE-Step 1.5 | 2.00 | 3.00 | 3.29 | Won the vocal-chop brief; used for chops and tempo-locked beats |
| MiniMax Music 2.6 | 2.17 | 2.71 | 2.86 | Retired after the test |
Yours to keepFiles and rights
Your files. No download counter.
Download as often as you like
Your plan sets how much you can generate. It never limits what you take home.
WAV and stems
MP3 and FLAC on every plan. WAV export and separated stems on Creator.
A commercial licence
Every paid plan lets you release and monetise what you make. The plain-English version is on the music rights page.
Labelled the way the rules ask
The EU AI Act asks for AI-made audio to be marked, and platforms are following. Lyro writes an AI-disclosure tag into every download, and Lyria tracks carry Google’s SynthID watermark.
The dealOn every plan
Fair terms, written down.
A failed generation costs nothing
If a track does not come out, every credit for it goes back by itself, and the app tells you.
Unused credits carry over
What you do not use this month moves to the next one, up to a month’s worth.
Cancel in two clicks
Inside the app, no email, no phone call. Or pause for up to three months instead.
Downloads are never counted
MP3, FLAC and WAV, as often as you like. What you make is yours to use commercially.
Nothing is deleted behind your back
If a plan ends, your library stays for 30 days and we remind you before anything goes.
Questions
What is Lyro?
Lyro is an AI music studio that turns a description, a vocal or a beat into a finished, mixed and mastered song. It runs in the browser, uses more than one music model and picks between them by task (Google Lyria 3 Pro for sung lyrics, ACE-Step 1.5 for vocal chops and tempo-locked beats), and includes a lyric writer, a stem splitter, a multitrack mixer and mastering. Lyro is operated by RNIFY Limited, a UK company.
Do I need to know how to make music?
No. You can start from a sentence such as “a warm afro house track for a sunset video, female vocal in Spanish”. If you do produce, you can set the tempo, key, lyrics and model yourself, then take the stems into your own DAW.
Can I use my own vocal or my own beat?
Yes. Upload or record a vocal and Lyro writes a beat around it. Upload a beat and Lyro adds a vocal, either singing your lyrics at the beat’s tempo and key or improvising a topline. If you have both, open them in the mixer and master the result.
Which languages can it sing in?
English, Turkish, Spanish, Portuguese, French, German, Italian, Arabic, Hindi, Japanese, Korean and Russian. Quality differs between models and languages. Turkish is the only language we have run a listening test in so far: the model Lyro uses for sung Turkish scored 4.67 out of 5 for pronunciation in that test (one rater, seven briefs). For other languages, listen to a first result before you commit to a long session.
Can I release songs made on Lyro on Spotify, TikTok or YouTube?
Paid plans include a commercial licence, so you may release and monetise your tracks. Each platform and distributor has its own rules about AI music and some distributors do not accept it, so check before you upload. Our guide Can you monetise AI music? keeps a current list.
Who owns the music I make?
As between you and us, it is yours: we claim no rights in your tracks and we license you everything we can. What we cannot promise is copyright protection itself. In the United States, music generated entirely by AI is not protected by copyright, and UK law on computer-generated works is under review. The details are on the music rights page.
How much does it cost?
Lyro is a subscription with monthly credits. There is a small plan for getting started, a main plan you size with a slider, and a custom plan for teams and agencies. You see the exact price before you pay and you can cancel online at any time. See the plans.
Is there a free plan?
No. Every generation costs us real money at the model provider, and we would rather spend it on better models than on free credits. You can listen to real output on this page, use the free prompt generator, and cancel whenever you like.
How is Lyro different from Suno or Udio?
Three things, and they are narrower than you might expect. Lyro does not count your downloads. Its mixer and mastering are included on Creator and use no credits. And it chose its model for sung lyrics with a small Turkish pronunciation test. Starting from your own vocal or beat is not a difference: Suno and Udio take uploaded audio too. Both of those products are strong at what they do, and both cost less per song; our comparison pages list the differences with sources.
Your turn.
A few taps about what you want to hear, then your first track.