AI song maker: describe a song, hear it sung

An AI song maker, also called a song maker AI or music maker AI, composes, arranges and sings a new song from a written description. Instead of recording instruments and vocals, you describe what you want to hear, and a music model generates the finished audio: melody, harmony, rhythm, production and voice in one pass.

AI Song Maker runs four of these models for you in the browser: Google's Lyria 3 Pro and Lyria 3, ElevenLabs Music, and MiniMax Music for instrumentals. There is nothing to install, no studio or microphone to set up, and no music theory to learn first.

It is a generator, not a full studio: you get one mixed MP3, not separate stems or a multitrack project. To cut, fade or master a song, open the MP3 in any audio editor.

Each model makes a different kind of track:

• Lyria 3 Pro (20 credits): full songs of up to about 3 minutes with verses, choruses and bridges. The default for finished songs.
• Lyria 3 (10 credits): 30-second clips with vocals, for hooks, jingles, intros and testing a prompt cheaply.
• ElevenLabs Music (120 credits): 60-second tracks with vocals or instrumental, with a different sound from Lyria.
• Instrumental, MiniMax Music (10 credits): music that never has vocals, for videos, podcasts, games and study playlists.

Music models respond best to the words musicians use. Cover these, and leave out anything you do not care about:

• Genre and era: 90s boom bap, modern country pop, 70s disco, synthwave
• Mood and energy: wistful, triumphant, playful, slow-burning
• Tempo: slow, mid-tempo, or a number such as 92 BPM
• Instruments: fingerpicked acoustic guitar, 808 bass, string quartet, brass stabs
• Vocals: female or male, range and texture such as smoky alto or bright tenor, a duet or a choir
• Topic and story: who is singing, to whom, and what happens

Put each part's tag on its own line, such as [Verse], [Pre-Chorus], [Chorus], [Bridge] or [Outro], and keep lines short enough to sing in one breath. Repeat the chorus word for word where you want it to return; that repetition is what makes a song stick.

Match the lyrics to the length. A 30-second Lyria 3 clip has room for a verse or a chorus, not a whole song. A Lyria 3 Pro song of up to about three minutes fits two verses, a repeated chorus and a bridge. The description and lyrics together can be up to 3,000 characters.

Every song here is newly generated, not a cover or a voice clone. Describe a style by genre, era and instrumentation rather than asking for a named artist's voice, and only paste lyrics you wrote or have the right to use. Google's Lyria models block requests for specific artist voices and for copyrighted lyrics; a blocked song fails and its credits are refunded.

Lyria 3 and Lyria 3 Pro add SynthID, Google's inaudible watermark, to every track so it can be identified as AI-generated. You may use your songs in personal and commercial projects under our terms and the model providers' terms, and we do not claim ownership of them. We cannot promise that an AI-generated song can be registered for copyright: the US Copyright Office's January 2025 report says prompts alone do not make you the author of AI output, while lyrics you wrote yourself remain your own work.