HeartMuLa is an open-source AI music generator that transforms text prompts, lyrics, or reference audio into production-ready songs, vocals, and instrumentals in seconds, with studio-quality output and fine-grained style control.
What is HeartMuLa?
HeartMuLa is an open-source music foundation model (4B parameters) that converts natural language descriptions, custom lyrics, or uploaded audio into complete music tracks. It runs on a web interface and supports five languages (English, Chinese, Japanese, Korean, Spanish). Powered by HeartCodec, it outputs 48kHz high-fidelity audio suitable for commercial use. The project is developed by the HeartMuLa team and hosted on GitHub with full model weights and training code available.
Key Features
- Text-to-Music Generation — Describe your desired music in plain language (genre, mood, tempo, instruments) and HeartMuLa creates a full track.
- Custom Lyrics with Vocals — Write your own lyrics and the AI generates a produced song with sung vocals, supporting multiple languages.
- Style Control — Precisely adjust genre, mood, tempo, and instrumentation through text prompts or reference audio.
- High-Fidelity Audio — HeartCodec encoding delivers 48kHz studio-quality output, as demonstrated in the showcase tracks (e.g., “Midnight Tokyo Drift,” “Electric Pulse”).
- Fast Generation — Complete tracks are produced in seconds using an optimized inference pipeline.
- Commercial Use — Generated music can be used in commercial projects (videos, games, ads) without copyright concerns.
- API Access — Integrate HeartMuLa into your own applications via a developer-friendly API, enabling real-time soundtrack generation.
- Self-Hosting — Deploy the open-source model on your own infrastructure for full data privacy and control.
Who is it for?
- Content creators and social media managers — Generate unique background music for YouTube videos, TikTok, Instagram Reels, and podcasts, avoiding copyright issues.
- Indie musicians and producers — Create song ideas, backing tracks, or complete productions, experimenting with different styles and arrangements.
- Game developers and app creators — Produce custom soundtracks and ambient audio for games and interactive experiences without licensing costs.
- Marketers and advertisers — Craft bespoke music for commercials, brand videos, and campaigns that match specific messaging and tone.
What can you do with HeartMuLa?
- Generate complete songs from text prompts — For example, input “Futuristic healing jazz, Japanese night city vibe, 90 BPM” to produce a track like “Midnight Tokyo Drift.”
- Produce custom backing tracks with vocals — Write your own lyrics in English, Chinese, or other supported languages, and HeartMuLa sings them with generated melodies.
- Create soundtracks for games or apps — Use the API to dynamically generate music that adapts to gameplay or user interactions in real-time.
How does HeartMuLa work?
- Describe your music idea – Specify style, mood, tempo, vocal preference, and instruments in the text prompt field (up to 500 characters).
- Generate – Click the “GENERATE” button and HeartMuLa produces a professional-quality draft in seconds (costs 2 credits per generation).
- Iterate with better prompts – Refine structure, energy, and arrangement by adjusting your description and regenerating.
- Export and use – Download the high-quality audio file (48kHz) for video, demo, or commercial projects.
