Public betaStable Audio 2.5 now in the generator

Hear your ideas.
In seconds.

Generate music, loops and SFX from a prompt, arrange them on a real multi-track timeline, and direct an AI Copilot that knows your mix. One project, from first idea to final master.

Start creating freeSee how it works

100 free credits · No card needed · Browser, Mac & Windows

soundgen.ai/studioExport WAV

Generate

Prompt

Model
ElevenLabs Music
Length
30s · stereo
Cost
20 credits

The models behind each feature, credited to the teams that built them.

  • ElevenLabs Musicsongs + vocals · ElevenLabs
  • Stable Audio 2.5instrumental + sfx · Stability AI
  • HTDemucsstem separation · Meta
  • Basic Pitchaudio → midi · Spotify
  • Claudecopilot · Anthropic

How it works

Prompt in.
Finished track out.

Every stage lives in the same project, so the sound you generate is the clip you arrange and the stem you export.

01Describe it

  • A vibe, a genre, a BPM, a key — or lyrics for a vocal
  • Full tracks, loops or sound effects
  • About 20 credits for 30 seconds of audio

02Arrange it

  • Generations land on a multi-track audio + MIDI timeline
  • Split any mix into up to 6 stems and keep the part you want
  • Or ask the Copilot — every action it takes has its own undo

03Finish it

  • Mix with busses, automation and a full effects rack
  • Master and export WAV up to 96 kHz / 32-bit
  • Bounce stems or MIDI, or share a link with playback
0kHz

Render quality

Studio-grade offline rendering, up to 32-bit float exports.

0+

Copilot tools

The AI agent operates the timeline, mixer, MIDI, and arrangement.

0

Built-in effects

EQ, compression, reverb, delay, and more on every track.

0

Stems per track

Deep separation splits a mix into vocals, drums, bass, guitar, piano, and everything else.

What's inside

Everything between
a prompt and a release.

Generation, arrangement, mixing and mastering share one project — nothing gets lost between a generator, a DAW and five browser tabs.

Describe a full track, a loop or a sound effect and studio audio lands in seconds — with vocals from lyrics, or instrumental. Chord detection, melody continuation and reference matching included.

Trust

Built to be trusted
with your music.

An AI studio is only useful if you can hand it your work. These are the defaults, not settings you have to find.

Private until you share it

New tracks and projects are private by default. A share link or the public player only goes live when you publish it.

Never trained on your audio

SoundGen does not train AI models on your audio, your prompts or your projects. Each request goes to a provider only to fulfil that request.

Yours to release

Commercial use on every credit pack. ElevenLabs output excludes film, TV, radio and released games; Stable Audio output has no such exclusion.

Every AI edit can be undone

The Copilot shows its plan before it touches your session, and every action it takes has its own undo.

Platforms

Start in the browser.
Keep going on the desktop.

One account across all of them — sign in anywhere and pick up the same projects.

Available

Browser

The full studio in any modern browser — nothing to install.

Open the studio
Signed & notarized

macOS

A native app with the same account — your projects are where you left them.

Download
Unsigned for now

Windows

Windows 10 and 11. SmartScreen may warn on first launch until the build is signed.

Download
Available

API & MCP

A REST API and an MCP server, so your own tools and AI assistants can drive SoundGen.

Read the docs

Why SoundGen

Most AI music tools stop
at the download.

A generator hands you a file. SoundGen keeps going — into the studio where the track actually gets finished.

Prompt-only generatorSoundGen
Download a stereo file, then start over in another appGenerations land on a multi-track timeline, ready to arrange
Re-roll the whole song to fix one partSplit it into up to 6 stems and edit the part that's wrong
The mix you get is the mix you keepA mixer with 49 effects, automation and busses — or ask the Copilot
One model, take it or leave itElevenLabs Music or Stable Audio, chosen per generation
A dead end for your existing sessionsOpen Ableton Live, REAPER, DAWproject projects; export stems and MIDI back out
A subscription to keep accessCredit packs, no subscription — 100 free to start

Who it's for

Made for people
who finish tracks.

Songwriters, producers, game developers and podcasters — anyone who needs the music to be done, not just started.

Start creating freeAll features
01

Songwriters

Hum an idea into your phone, describe the rest in a prompt, and have a produced demo before the feeling fades. Ask the Copilot to mix it while you stay in the writing.

Demos
02

Producers

Pull the drums out of an old live recording, drop them on the timeline as their own track, and rebuild the arrangement around them.

Stem separation
03

Game developers

Score a jam entry over a weekend — loops, stingers and ambient beds, described in words and exported as stems your engine can layer.

Loops & SFX
04

Podcasters

Make your intro, transitions and bed music in one place, then bounce the episode at 96kHz without opening a second app.

Ambient
05

Bedroom producers

Say "tighten the low end and widen the chorus" and watch the agent work the actual faders — with a Revert on every action it takes.

AI Copilot
06

Anyone mid-project

Open an Ableton or REAPER session with its arrangement intact, finish it in the browser, and export stems back out to your DAW.

DAW import

Pricing

Pay for what
you generate.

Generation is metered by length — about 20 credits for 30 seconds, 38 for a minute. Buy credits when you need them; there's no subscription.

Free

Try everything first

$0

100 credits when you sign up

  • No card required
  • The full studio — AI features draw on credits
Start free

Creator

For regular creators

Most popular

$3.99one-time

500 credits

  • 500 AI generation credits
  • ~13 minutes of generated audio
  • All AI models access
  • Stem separation
  • Commercial use of everything you generate
  • API & MCP access
Get started

Pro

For power users

$8.99one-time

1,200 credits

  • 1,200 AI generation credits
  • ~31 minutes of generated audio
  • All AI models access
  • Stem separation
  • Commercial use of everything you generate
  • API & MCP access
Get started
See full pricing and the credits FAQ

Changelog

Recently shipped.

Every release

2026.09.24 · September 24, 2026

Stable Audio 2.5 in the Generator

A second music model. The Generate page now lets you choose Stable Audio 2.5 alongside ElevenLabs Music for each generation — instrumentals and sound design up to 190 seconds, delivered as WAV rather than compressed audio

2026.09.07 · September 7, 2026

Take Lanes, Swipe Comping, and the Rest of Tracking

Comp by hand as well as by swipe. In the take lanes, double-click splits the section under the pointer, dragging a seam moves it (both sides trim together), and a click hands the section under it to that lane’s take. The lane’s name still means “the whole clip”

2026.09.06 · September 6, 2026

The Mac App Records Through Its Own Engine

In the native Mac app, recording now happens inside the audio engine. The input is opened by the engine itself, so every recorded block is stamped with the exact frame of the song it played against — takes land on the sample, punch in and out land on the sample, and a loop wrap starts a new take file on the boundary. The app used to record through the web page on a separate clock, placing takes to within a frame of video rather than a frame of audio

FAQ

Questions,
answered.

Can't find what you need? Get in touch.

Your next track
starts as a sentence.

Your first 100 credits are on us — that's 5 thirty-second generations to start with. No card required.

Start creating freeDownload for Mac & Windows