What is waivepulse?

WAIvePulse is a local, fully offline AI music generation tool. You give it structured lyrics using section markers like [Verse], [Chorus], and [Bridge], plus a set of style and genre tags, and it produces a complete song with vocals as an MP3, all running on your own machine with no cloud service or subscription. The tool serves a web interface at http://localhost:7860. When you submit a request, it works through two phases: the HeartMuLa 3B language model generates audio tokens from your lyrics and tags, then a codec converts those tokens into an audio waveform. You can queue multiple songs while one is generating. Each completed card shows the detected BPM and musical key, an audio player with several visualizer styles, and a download button. A built-in Studio page lets you separate any generated song into six stems: vocals, drums, bass, guitar, piano, and other, and mix them independently. You can mute, solo, and adjust volume per stem, or use one-click presets like Karaoke (vocals off) or Acappella (instruments off). Tags are the main way to shape the sound. The model was trained with eight categories: Genre, Timbre, Gender, Mood, Instrument, Scene, Region, and Topic. The key rule is one tag per category: stacking multiple genre tags produces muddier output. Setup downloads the model weights automatically (about 21 GB total). Requirements include Python 3.10 or newer, an NVIDIA GPU with roughly 10 GB of VRAM (CUDA required), and either Windows or Linux, macOS is not supported. The full README is longer than what was provided.