Free, local AI music studio · runs on your GPU

Your studio.
Your GPU.
Your music.

Write it. Generate it. Mix it. Loop it. Perform it. One app, five tools, free and offline — every note stays on your machine.

Free & open source No cloud No subscription No API keys No usage caps Runs on your GPU
01 Write
02 Generate
03 Separate
04 Mix
05 Loop
06 Perform
Generate HeartMuLa 3B

Type lyrics. Get a finished song.

Paste your words, pick a few style tags, and a 3-billion-parameter model writes, sings, and produces a complete track as an MP3 — vocals and all. Jobs queue up; BPM and key are detected automatically.

Generate screenshot
Lyrics Local Llama via Ollama

Writer’s block, solved — privately.

Stuck on a line? A local Llama model streams lyrics word-by-word, pre-formatted with the verse and chorus markers the generator expects. Then polish them: a syllable gutter flags lines that break the flow, rhyme schemes are colour-coded, and a built-in rhyme finder works fully offline.

Lyrics screenshot
Studio Demucs stem mixer

A full DAW in your browser.

Every song splits into six stems on your GPU — vocals, drums, bass, guitar, piano, other. Mix them with per-track EQ, volume and pan automation, a master chain, and a rack of creative effects. Match your master to any reference song, then export at YouTube loudness as WAV or MP3.

Studio screenshot
Looper Ed Sheeran in a tab

Build a track, layer by layer.

A full loop station: synthesized drums with a 16-step sequencer, a two-octave synth, a polyphonic piano roll, a plucked-guitar voice, a sampler, and a mic with autotune — stacked into six loops. Every note can be played by up to seven slightly detuned voices spread across the stereo field, so a part sounds like a section rather than a keyboard. Draw a dynamics curve under the roll and held notes get brighter as well as louder when they swell, then place each loop near or far in a real room.

Looper screenshot
Karaoke Performance mode

Lyric videos, ready to record.

Whisper transcribes your vocals and aligns them to your original words. A fullscreen visualizer rolls the lyrics over your tuned mix — record it straight to video, and export YouTube-ready captions to go with it.

Karaoke screenshot

Why WAIvePulse

A one-time install in exchange for permanent freedom from the cloud.

WAIvePulse Cloud generators Pro DAWs
Runs offline ✓ — ✓
Open source ✓ — —
Songs from lyrics ✓ ✓ —
Browser stem mixer ✓ — —
Built-in loop station ✓ — add-on
Lyric-video output ✓ — —
YouTube-loudness export ✓ — add-on
Reference-track mastering ✓ — add-on
Finds the flaws in a render for you ✓ — add-on
Writes a new instrument into your song ✓ — —
Trains on your own songs, locally ✓ paid —
Rewrite one section of a finished song ✓ paid —
Webcam instrument ✓ — —
Pronunciation respelling for lyrics ✓ — —
MIDI keyboard input ✓ — ✓
Ensemble + expression synth ✓ — add-on
Album art for every song ✓ — —
Subscription None Monthly Monthly
Usage caps None Credits None
6stems per song
5tools, one app
0API keys needed
∞songs, no limits

Using it at work

Almost everything is free. One thing isn’t, and it is easy to tell apart.

Free, forever, no permission needed

  • Use it yourself, at home or at work
  • Sell the songs you make with it. The models are Apache 2.0 and MIT; nothing claims your output
  • Use your songs in client work, ads, games, videos
  • Modify it, fork it, run it on every machine in your studio
  • Teach with it, stream it, write about it

Email first

  • Selling WAIvePulse, or a product whose value is substantially this software
  • Running it as a paid service for other people
  • Charging for hosting, support or installation of it
  • Shipping it inside something you sell

A commercial licence covers it. Indicative: $149/yr for a one-person studio, $499/yr for a team or an embedded product. Say what you are building and I will tell you honestly whether you even need one.

Ask about a licence

Make music the cloud can’t touch.

Free and open source. Windows or Linux, an NVIDIA GPU with ~12 GB, and you’re running your own studio in minutes.