AI songwriting pipeline — music with soul
Describe a song, paste a YouTube URL, or upload audio to extract metadata for your next track.
Tell me what you're going through, or browse the archetypes below.
Click = primary. Click another = modifier.
Be honest — the more specific and real you are, the better the song.
Shape the sound. Accept the AI suggestions, tweak them, or skip this step entirely.
Describe a song, paste a YouTube URL, or upload audio. The extracted style will be applied to your current metadata.
The emotional architecture — when to whisper, when to roar.
Format: Label | bars | energy 0-1 | density —
density: none, sparse, moderate, dense
This is where your song comes alive.
Two separate outputs: an instrumental backing track, and a full song with vocals.
Generates a backing track without vocals. Use this for karaoke-style playback or to hear the instrumental arrangement.
Generates a complete song with AI-sung vocals. This is a separate output from the instrumental above.
Band voice config, genre & metadata are included automatically. Override here if needed.
Generates a full song (vocals + instrumental) in one pass using MiniMax Music 2.6. Supports up to 3500 chars of lyrics with rich section tags, plus a 2000-char style prompt. Different model from MusicGen + ACE-Step above — try both and compare!
Up to 3 minutes of high-quality instrumental music. No lyrics — purely instrumental. Great for longer backing tracks.
Take an existing song and re-create it in a completely different style. "What if this was jazz?" "Make it lo-fi." "Acoustic stripped version."