AI MUSIC GENERATOR

AI Music Generator

Describe the song you want — genre, mood, instruments, tempo, subject — and the generator returns a finished track with vocals or an instrumental arrangement. One song costs 5 credits, and a new account starts with 20.

Describe a song⌘ + ⏎
0 / 3000
Style tagsOptional · pick up to 5
Not sure? Start from one of these

ABOUT THIS TOOL

Turn a Text Prompt into a Finished Song

AI music generation starts from words rather than a piano roll. You describe the song; the generator writes the arrangement and the performance and hands back audio you can play in the browser. From there the same song can be continued, partly rewritten, or blended into another track — each a separate action that bills against your credits.

Output
A generated song in your workspace, playable in the browser and downloadable as a WAV for 1 credit. The song keeps the prompt, model and settings that produced it, so Extend, Replace section and Mashup each build on it as a separate Generation while the original stays playable.

How it works

  1. 01Write a prompt: genre, mood, instruments, tempo, and what the song should be about.
  2. 02Stay in Simple mode for prompt-and-go, or switch to Advanced mode to write the lyrics yourself and set style, title, and vocal gender.
  3. 03Generate the track. It lands in the workspace panel, where you can play it, download it, or send it to Extend, Replace section, or Mashup.

Key features

  • Text-to-song generation with an instrumental switch
  • Simple and Advanced modes over the same generator
  • Model options from V6 Mini up to V6 Wild
  • Every result stays in the workspace, ready for Extend, Replace section, or Mashup

Settings explained

Simple
Simple or Advanced. Simple takes a description of the song plus an optional instrumental switch; Advanced exposes the lyrics, style tags, vocal gender, weirdness and style influence, and adds a title. Both write the same request, so switching between them keeps what you have already typed.
Model
Which model writes the track. V6 aims for greater musical expression with more natural vocals and richer detail; V6 Wild pushes for bolder, less predictable results; V6 Mini trades some quality for speed. The choice is saved with the song, and Extend inherits it from the source rather than asking again.
Vocal Gender
Vocal gender, or None to let the model decide. It applies only when the song has vocals, so it does nothing while the instrumental switch is on. It shapes the performance, not the key or the tempo.

What each action costs, and where it runs

Generating a song costs 5 credits, and Extend and Mashup cost 5 each. Replace section costs 3 and downloading the WAV costs 1. Every action that creates or exports audio is a cloud action: it needs a signed-in account, bills credits, and sends your prompt or your audio to a remote generation service. Playing, pausing and seeking through a finished song runs in your browser and costs nothing. A new account starts with 20 credits, which is four generated songs, and a separate daily limit of twenty generations per account caps how often the service can be triggered — an abuse backstop rather than a pricing tier, so credits run out first.

What the generator does with your words

AI music generation starts from language, not a piano roll. You describe the song — genre, mood, instruments, tempo and subject — and the model writes the arrangement and the performance together, then returns audio. That is why the same prompt can come back with a different melody each run: the model is interpreting a description, not retrieving a recording. The controls sit on top of that interpretation. Style tags are read as concrete attributes such as 'warm analogue keys' or 'slow tempo', which is the cheapest way to steer a result; the model choice decides how far the model is allowed to stray; and vocal gender shapes the performance when the song has vocals. None of them fix the exact notes, so treat the first result as a direction and the controls as the way to move it.

From a finished song to a finished edit

A generated song is a source, not an end state. It can be downloaded as a WAV for 1 credit, extended past its ending, have one passage of 6 to 60 seconds rewritten, or be blended with another track in a mashup. Each of those produces a new Generation in the same workspace rather than modifying the original, so a version you liked is never overwritten by the next experiment. Because the source parameters travel with the song, Extend continues from the model and settings that produced it instead of asking you to describe the track again. Uploads can join the workspace as playable songs and be used as mashup sources; they cannot be extended or partly rewritten, because those actions need the identifiers of a generated song.

Common questions

Yes. Generating, extending, replacing and mashing all bill credits against your account, so the generator asks you to sign in before it submits a job. You can browse this page and play the showcase tracks without signing in.