Text to music
Write what the music should sound like and what it is for. Get a song with vocals or a clean instrumental back as an MP3, made by the model you choose.
AI Song Maker is a music generator AI that turns text into music. Describe the genre, mood, tempo, instruments and vocals; the model composes and produces the whole track and returns an MP3. Choose Lyria 3 Pro for songs of up to about 3 minutes (20 credits), Lyria 3 for 30-second clips (10), ElevenLabs Music for 60-second tracks (120) or MiniMax Music for instrumentals (10). New accounts get two free Lyria 3 Pro songs (40 credits).
Pop, rock, hip-hop, country, jazz, EDM, lo-fi, orchestral and the sub-genres in between. Name it the way a musician would and the model follows.
Ask for a singer and the singing models write and perform lyrics, or perform the lyrics you paste. Ask for no vocals and you get an instrumental.
The MiniMax instrumental model never sings, so beds for videos, podcasts and games stay clear under a voiceover.
Lyria 3 for 30-second clips, ElevenLabs Music for one minute, Lyria 3 Pro for songs of up to about three minutes.
Each model has one price per track, shown before you generate. Tracks that fail are refunded automatically.
Play every track in the page, download the MP3, and find it again later in your history.
Genre, mood, tempo, instruments and vocals in one or two sentences. Add what the music is for, such as a trailer or a study playlist.
Lyria 3 Pro for a full song, Lyria 3 for a short clip, ElevenLabs Music for a minute, or the instrumental model for no vocals.
Listen, download, or change one thing in the description, such as the tempo or the singer, and generate again.
Whether you call it a music AI generator or an AI music generator, a text-to-music model learns how descriptions relate to recorded sound. Given your words, it generates the audio directly: the arrangement, the performance, the mix and, on singing models, a vocal line with lyrics.
That is why the description matters so much. The model only knows what you tell it, so vague prompts get generic music and specific prompts get specific music.
A reliable pattern is genre, mood, tempo, instruments, vocals and purpose, in that order. For example:
Not a real one. Every track takes serious computing power, so free plans always have limits somewhere: on quality, length, downloads or commercial use. AI Song Maker is upfront instead: new accounts get two free Lyria 3 Pro songs (40 credits), and after that each track has a fixed price you see before you generate.
The cheapest way to explore is Lyria 3 at 10 credits per 30-second clip, about $0.09–$0.12. When a clip sounds right, make the full song with Lyria 3 Pro.
Your two free Lyria 3 Pro songs include vocals: describe the singer in your prompt, then paste lyrics or let Lyria write them. Google lists English, German, Spanish, French, Hindi, Japanese, Korean and Portuguese as Lyria 3's vocal languages.
Prices are per finished track, whatever you describe.
| Lyria 3 Pro | Lyria 3 | ElevenLabs Music | Instrumental | |
|---|---|---|---|---|
| Vocals | Yes | Yes | Yes | Never |
| Length | Up to about 3 min | 30 s | 60 s | Set by the model |
| Credits per track | 20 | 10 | 120 | 10 |
| Best for | Finished songs | Hooks, jingles, tests | One-minute tracks | Background music |
Yes. A text description is all the models need. Add lyrics if you want specific words sung; otherwise the singing models write their own, and the instrumental model makes music without words.
New accounts get two free Lyria 3 Pro songs (40 credits). It is not unlimited: after the free songs, each track is paid with credits, from 10 credits for a 30-second Lyria 3 clip.
Any genre you can describe, from pop and hip-hop to orchestral and ambient. Niche styles work best when you also name the instruments and tempo, because that tells the model what the genre sounds like to you.
Yes, in personal and commercial videos, under our terms and the model providers' terms. We do not claim ownership of your tracks. Follow each platform's rules on AI-generated content.
Describe it in a sentence and hear it back as a finished track.
Last updated October 6, 2026