HeartMuLa ❤️

The best open-sourced music generation model in terms of lyrics controllability and music quality

Super-fast torch compiled version

Generate music from lyrics and tags using HeartMuLa, an open-source music foundation model.

Instructions:

Enter your lyrics with structure tags like [Verse], [Chorus], [Bridge], etc.
Add comma-separated tags describing the music style (e.g., piano,happy,romantic)
For best results use one tag per category
Adjust generation parameters as needed
Click "Generate Music" and wait for your song!

Note: Generation can take several minutes depending on the duration.

Lyrics

[Intro]

[Verse]
Frustrated, degraded
Down before you're done
Rejection, depression
Can't get what you want

[Prechorus]
You ask me how I make my way
You ask me everywhere and why
You hang on every word I say
But the truth is like a lie

[Chorus]
Live to win 'til you die
'Til the light goes out in your eyes
Live to win, take it all
Just keep fighting 'til you fall

[Verse]
Obsessive, compulsive
Suffocate your mind
Confusion, delusions
Kill your dreams in time

[Prechorus]
You ask me how I took the pain
Crawled up from my lowest low
Step by step and day by day
'Til there's one last breath to go

[Chorus]
Live to win 'til you die
'Til the light goes out in your eyes
Live to win, take it all
Just keep fighting 'til you fall
Day by day, kickin' all the way
I'm not caving in
Let another round begin, live to win
Yeah, live, yeah, win

[Outro]

Tips for Better Results

Use structured lyrics with section tags
Be specific with your style tags and use one tag per style category
Try different temperature values for variety
Shorter durations generate faster

Offical Tags

GENRE (95% - MANDATORY) Examples: Pop, Rock, Electronic, Hiphop, Jazz, Classical, Techno, Trance, Ambient.
TIMBRE (50% - Sound Texture) Examples: Soft, Warm, Husky, Bright, Dark, Distorted.
GENDER (37% - Vocal Character) Examples: Male, Female.
MOOD (32% - Emotional Vibe) Examples: Happy, Sad, Energetic, Joyful, Melancholic, Relaxing, Dark.
INSTRUMENT (25% - Dominant Sounds) Examples: Piano, Synthesizer, Acoustic Guitar, Electric Guitar, Bass, Drums, Strings, Violin.
SCENE (20% - Listening Context) Examples: Dance, Workout, Dating, Study, Cinematic, Party.
REGION (12% - Cultural Influence) Examples: K-pop, Latin, Western.
TOPIC (10% - Lyrical Theme) Examples: Love, Summer, Heartbreak.

Model: HeartMuLa-oss-3B | Paper: arXiv | Code: GitHub

Licensed under Apache 2.0