music-generation

name	music-generation
description	AI music generation via Replicate — 5 models for background tracks, lyrics, and sound design
tier	extended
applyTo	**/*music*,**/*soundtrack*,**/*audio*,**/*score*
$schema	../SKILL-SCHEMA.json

Music Generation Skill

Domain: AI Music Generation
Version: 1.0.0
Last Updated: 2026-04-15
Author: Alex (Master Alex)
Source: Patterns from AlexVideos CLI toolkit
Related: text-to-speech (voice), video-generation (video)

Overview

Generate AI music and soundtracks via Replicate. 5 models covering lyrics-based songs, instrumental tracks, and professional sound design.

---

Model Catalog (5 Models)

Key	Model	Replicate ID	Cost	Features
music15	MiniMax Music 1.5	minimax/music-1.5	$0.03/file	Lyrics with [verse], [chorus] tags
music01	MiniMax Music 01	minimax/music-01	$0.035/file	Reference audio input
stableaudio	Stable Audio 2.5	stability-ai/stable-audio-2.5	$0.20/file	Diffusion-based, 1–190s
elevenmusic	ElevenLabs Music	elevenlabs/music	$8.30/1k sec	Professional, vocals toggle
lyria2	Google Lyria 2	google/lyria-2	$2/1k sec	Negative prompt support

---

Parameter Support Matrix

Model	prompt	lyrics	duration	reference	negative	vocals
music15	✅	✅	—	—	—	✅ auto
music01	✅	—	—	✅	—	—
stableaudio	✅	—	1–190s	—	✅	—
elevenmusic	✅	—	variable	—	—	✅ toggle
lyria2	✅	—	variable	—	✅	—

---

Model Selection Guide

Need	Model	Why
Song with lyrics	music15	Structured lyrics with verse/chorus
Match existing style	music01	Reference audio input
Budget instrumental	stableaudio	Cheapest per-file
Professional production	elevenmusic	Highest quality, vocal toggle
Exclude unwanted elements	lyria2, stableaudio	Negative prompt support

---

Lyrics Format (MiniMax Music 1.5)

Use structured tags for song sections:

[intro]
(Soft piano melody)

[verse]
Walking through the city lights at night
Every shadow tells a different story
Searching for a meaning in the dark
Finding hope in unexpected places

[chorus]
We rise above the noise
Breaking through the silence
Every heart beating as one
Together we are stronger

[bridge]
When the world falls apart around us
We'll find our way back home

[outro]
(Fade out with gentle strings)

---

Generation Examples

Song with Lyrics (MiniMax)

const output = await replicate.run("minimax/music-1.5", {
  input: {
    prompt: "Uplifting pop rock song about finding hope",
    lyrics: `
[verse]
Walking through the rain...
[chorus]
We are the dreamers...
    `,
  },
});

Instrumental (Stable Audio)

const output = await replicate.run("stability-ai/stable-audio-2.5", {
  input: {
    prompt: "Epic cinematic orchestral score, heroic theme, brass and strings",
    duration: 60, // seconds
    negative_prompt: "vocals, singing, speech, drums",
  },
});

Reference-Based (MiniMax 01)

const output = await replicate.run("minimax/music-01", {
  input: {
    prompt: "Upbeat electronic dance track",
    reference_audio: referenceAudioDataURI, // Sample to match style
  },
});

Professional with Vocals Toggle (ElevenLabs)

const output = await replicate.run("elevenlabs/music", {
  input: {
    prompt: "Energetic pop song with catchy hook",
    include_vocals: true, // or false for instrumental
  },
});

---

Prompt Engineering

Element	Good Example	Avoid
Genre	"lo-fi hip hop", "orchestral cinematic"	"good music"
Mood	"melancholic", "uplifting", "tense"	"nice"
Instruments	"acoustic guitar, soft piano, strings"	"instruments"
Tempo	"slow ballad 60 BPM", "upbeat 120 BPM"	—
Structure	"builds to crescendo", "ambient loops"	—

---

Cost Comparison (60s track)

Model	Cost	Quality	Notes
stableaudio	~$0.20	Good	Budget, diffusion-based
music15	~$0.03	Good	Best for lyrics
music01	~$0.035	Good	Style matching
lyria2	~$0.12	High	Negative prompts
elevenmusic	~$0.50	Premium	Professional quality

---

Integration with Studio Agents

• Video Generation: Background music for generated videos
• Text-to-Speech: Combine speech with background tracks
• Video Editing: Merge music with video via avmerge

---

Common Use Cases

Use Case	Recommended Flow
YouTube video music	stableaudio → descriptive prompt → 60–120s
Song demo	music15 → structured lyrics → review
Brand audio logo	stableaudio → 5–10s → specific style prompt
Podcast intro	stableaudio → 15–30s → upbeat, no vocals
Film score	lyria2/elevenmusic → cinematic prompt → long form

---

Output Handling

import { writeFileSync } from "fs";

// Music typically returns MP3 or WAV
const response = await fetch(output);
const buffer = await response.arrayBuffer();
writeFileSync("output.mp3", Buffer.from(buffer));