About SoundFrost

SoundFrost generates original music with AI. You describe a style — genre, mood, instruments, tempo, the kind of voice you want — optionally add your own lyrics, and the engine performs a complete track, vocals included. Nothing in the catalog is a copy of an existing recording.

How it works

Behind the site sits an ACE-Step diffusion model running on a dedicated GPU. Your description becomes a conditioning prompt; the model writes the arrangement and sings the lyrics you supplied. A two-minute track takes roughly a minute of GPU time, which is why requests are queued rather than answered instantly.

The desktop app

The Windows app is the same catalog and the same generator in a native window, plus the things a browser cannot do well: a real folder of MP3s on your disk, offline playback, and a queue that survives closing the window. It is a single portable executable — no installer, no admin rights. Get it from the download page.

Using the music

Tracks are free to listen to and free to download. Because the audio is machine-generated from your own prompt, you are free to use it in videos, streams and personal projects. Two rules keep it that way: don't submit lyrics you don't own, and don't present a generated track as a recording by a real artist.

History

soundfrost.org has been around since the era when it was a music downloader for the desktop. The domain, and the desktop habit it came with, are now pointed at making music instead of fetching it.

Contact

Questions, bugs, takedown requests: [email protected].