ai-music · AI Music & BGM Generation

Generates original background music, scores, and instrumentals for short videos, vlogs, and social content: describe the style, mood, and instruments, and the skill submits → polls → downloads asynchronously, with optional trimming, loudness normalization, or mixing into video.

Supply lyrics for a vocal version; works with two providers — DashScope and Suno-compatible gateways. Bring your own API key; usage is metered.

Example invocation: "Generate a 30-second upbeat lo-fi instrumental for my vlog intro BGM."

Full brief

Positioning

ai-music is the text-to-music entry point: original background music, scores, and instrumentals for videos and social content. It does music only — no voiceover, no narration.

Core capabilities

Instrumental BGM generation: style / mood / instrument description → mp3; duration settable on supported providers.
Vocal mode: supplying lyrics switches to a sung version instead of an instrumental.
Pluggable providers: dashscope (Alibaba Bailian) / suno-compatible (third-party Suno-like gateways) — switching providers only changes --provider plus env vars; the flow stays the same.
Full async flow: submit → poll → download (poll interval and timeout adjustable).
Post-processing via shared scripts: trim to length, loudness-normalize to -14 LUFS, mix the BGM into video (adjustable voice/BGM volume ratio, auto-loop and truncate).

Workflow

Confirm the available provider: run model_registry.py configured --group music; if several are available and the user didn't name one, list them and ask
check the configuration (no request sent; missing keys get a clear Chinese prompt — no wasted failed calls)
generate the music
Optional post-processing: trim / normalize / mix into video

Inputs & outputs

Input	Notes
prompt	Required — style / mood / instrument description
provider	Optional — dashscope or suno-compatible (env override supported)
lyrics	Optional — triggers vocal mode when supplied
duration	Optional — seconds; supported by some providers
instrumental	Optional — instrumental-only BGM
output	Optional — defaults to `outputs/<topic>/{name}.mp3`

Output: music audio file (mp3), with the actual provider / model, polling trace, and final path printed.

Boundaries with adjacent skills

tts-voiceover: human voiceover / narration / read-aloud; ai-music (this): background music and scores. They compose: BGM + narration mixed via video_ops.py bgm.

Fit

Intro/outro BGM for short videos and vlogs
Scores for social content, podcast interstitials
Original instrumental beds where needed

Before you start

Configure keys in the project-root .env: dashscope needs DASHSCOPE_API_KEY; suno-compatible needs MUSIC_API_KEY + MUSIC_BASE_URL.
Metered billing: run check before generate; keys are filled in by the user — the skill never requests, echoes, uploads, or commits real keys.
Never overwrites source material; only writes new files under outputs/<topic>/.

ai-music is part of the Aiglade Skill library. Invoke it from the Aiglade chat box in plain language.

← All 155 skills