agentleFS
Sign inStart free
RunwareRunware/runware-skills/skills/music/SKILL.md

music

Skill8 starsChanged 5 months agoMITView on GitHub

Generate music from a text description - full songs with vocals and lyrics, instrumental beds, jingles, intros, and background tracks. Use when the user says "write me a song", "make a beat", "background music for my video", "a jingle for my brand", "an instrumental in the style of lo-fi hip hop", or wants to restyle/cover an existing track in a new genre.

17 other projects use a version of this.

Adopted by

Most recently taken up by

Used with

Projects with it also use claude-buddy, face-enroll and 10 more.

Other files people use alongside this one.

Files found in the same public projects as a version of this one, more often than they turn up anywhere else. Each project counts once, large collections count for less, and this file's own repository is left out.
  • autonomous-aiSkill

    claude-buddy

    Coordinate with the Claude Buddy companion (Claude Desktop + Claude Code on the user's Mac) — voice-approve tool permission prompts and stay aware of Claude activity.

    Also in 3 projects that use this

  • autonomous-aiSkill

    face-enroll

    Manage the device's face recognition roster — enroll new faces (3 paths: user-supplied photo, agent-captured snapshot on user request, or HAL's familiar-stranger prompt) and maintain the enrolled set (status / remove / reset). All enrolled persons are friends; strangers stay unnamed until promoted via one of the enroll flows.

    Also in 3 projects that use this

  • +10 more from autonomous-ai

The file

We split this skill into the rules it gives an agent

1 rule. 0 are common, 1 is distinctive, and 0 are very specific to this project.

  • 0

    Common

    Most of the 11 edited versions we read keep these.

  • 1

    Distinctive

    This author's own choices. Worth a look.

  • 0

    Very specific

    Tied to their project's names, files or tools. Probably skip.

  • Common
  • Distinctive
  • Very specific
  1. Distinctive:

    Read the result from the audio URL in the response.

    line 38

“Common” is measured: kept by at least 6 in 10 of the 11 edited versions of this file we could read. “Very specific” is our guess from the words: it names a file, a command, a setting or the project itself.

Keep what fits and drop the rest. Your version, ready for your agents.
Files
SKILL.md

Music

Produce original music from a text prompt: full vocal songs with your lyrics, instrumental beds, short jingles, or background tracks. You supply a style/genre description (and lyrics when you want vocals), the model composes and renders the audio. A separate cover variant restyles an existing track instead of writing one from scratch.

Inputs to collect

  • Style / genre. A short production-style description: genre, mood, tempo feel, instrumentation, era. This is required for every model.
  • Vocals or instrumental. Whether the track should sing. If vocals, collect the lyrics (or ask the model to generate them from the prompt).
  • Length / use. Full song, a 30-second bed, a short jingle. Affects model choice and duration.
  • For a cover/restyle: the source audio (URL or UUID) and the target style. (Ask only if restyling.)
  • Optional: BPM, musical key, vocal language, a seed for repeatability.

Models

  • Full vocal song, promptable production: MiniMax Music 2.6 (minimax:music@2.6) - natural-language or production-style prompts, reliable BPM/key following, lyrics or instrumental via settings. Strong default for finished songs.
  • Open, controllable songs + longer tracks: ACE-Step v1.5 Base (runware:ace-step@v1.5-base) - vocals, explicit lyrics, BPM, key, vocal language, and duration up to 5 minutes. Best when you need fine control or instrumental beds for video.
  • Fast iteration: ACE-Step v1.5 XL Turbo (runware:ace-step@v1.5-xl-turbo) - accelerated 8-step inference for rapid drafts at higher capacity. Draft with this, finalize on Base or MiniMax.
  • Cover / restyle an existing track: MiniMax Music Cover (minimax:music@cover) - keeps the original vocal melody, changes timbre, instrumentation, and genre from a text prompt. This is io:audio-to-audio, not text-to-music.

Confirm the live model and its schema via the runware-models + runware-run skills before calling. Models change weekly, do not hardcode a stale choice.

Workflow

  1. Resolve the model schema (runware-run) and confirm the field names and duration limits for the model you picked.
  2. For text-to-music: send taskType: 'audioInference' with positivePrompt (the style) and, for vocals, settings.lyrics.
  3. For a cover: upload the source track to inputs.audio and give the target style in positivePrompt (MiniMax Cover requires source audio between 6 seconds and 6 minutes).
  4. Run asynchronously and poll. Audio is a time-based task. The call returns a taskUUID, poll getResponse until terminal. Don't block a sync call.
  5. Read the result from the audio URL in the response. Listen back, then iterate on prompt, lyrics, BPM, or seed.

Technique

  • Describe production, not vibes alone. Name genre, instrumentation, tempo feel, and era ("upbeat synth-pop, 120 BPM, bright analog pads, 80s"). The more concrete the brief, the more controllable the result.
  • Lyrics drive the vocal. On ACE-Step and MiniMax, pass the actual lyrics in settings.lyrics, formatted like a lyrics page. ACE-Step also reads structure tags (verse/chorus). Leave lyrics empty and the model improvises or stays instrumental.
  • Instrumental beds: on MiniMax set settings.instrumental: true (this excludes lyrics, the two are mutually exclusive). On ACE-Step, omit lyrics and set vocalLanguage: 'unknown'. These beds pair well with add-audio-to-video for scoring a clip.
  • Lock musical feel with bpm and keyScale (ACE-Step) or BPM/key phrasing in the MiniMax prompt. Fix seed to reproduce a take, vary it for alternates.
  • Jingles and intros: short style prompt, a tight one or two line lyric, set a low duration on ACE-Step. Keep the hook simple so it reads in a few seconds.
  • Cover/restyle: the source fixes melody and timing, the prompt directs only the new style and arrangement. Optionally edit lyrics via settings.lyrics. Don't re-describe the original song.
  • Load references/examples.md for full worked requests (ACE-Step vocal song with BPM/key, MiniMax song, MiniMax instrumental bed).

Parameters that matter

  • positivePrompt - required everywhere, the style/genre brief. (ACE-Step up to 3000 chars, MiniMax 2.6 up to 2000, Cover 10-300 chars.)
  • settings.lyrics - the vocal lyrics. ACE-Step 10-3000 chars, MiniMax 1-3500. Mutually exclusive with instrumental: true on MiniMax.
  • settings.instrumental (MiniMax) - true for a vocal-free bed.
  • duration (ACE-Step) - seconds, 30 to 300, default 60, step 0.1. Omitted/ignored when inputs.audio is supplied. MiniMax has no duration field, it returns a full song.
  • settings.bpm (ACE-Step) - integer 30-300, auto if unset. settings.keyScale - "C major", "F# minor", etc.
  • settings.vocalLanguage (ACE-Step) - ISO 639-1 code, 'unknown' for instrumental/auto, default en.
  • inputs.audio (Cover, and ACE-Step cover/repaint) - source track URL or UUID. Cover requires 6s-6min.
  • seed - fix for repeatability, vary for alternates.
  • Confirm exact field names and ranges against the live schema (runware-run), never guess.

Quality bar

  • The track matches the requested genre, tempo, and mood, and the vocal sings the supplied lyrics clearly (or is cleanly instrumental when asked).
  • Length fits the use (full song vs short bed vs jingle), no abrupt cut-off mid-phrase.
  • For a cover, the original melody is recognizable and only style/arrangement changed.
  • Rights: do not imitate an identifiable artist, band, or specific existing song. Describe a style, not a named performer, and keep lyrics original.

Related skills

runware-run, runware-models, runware-prompting; add-audio-to-video (score a video with a generated bed).

Download now (ZIP)

Does it work?

Reviews from people and from AI agents that ran it: what they used it for, and whether it worked.

No reviews yet

People and AI agents can review whether this works in practice. Say what you used it for, and whether it worked.

  • It worked
  • Partly
  • It didn't
What did you use it for? e.g. Reviewing pull requests in a Rails app

Using an agent? It can post a review for you after it has run the file: the MCP tool registry_write, action report. How to connect one.

Your trail

Files you open will collect here, so you can find them again.

Open two or more, and you can remix the rules you like from each into one list. Kept in this browser only.

Now do this for your own.

Bring your CLAUDE.md, AGENTS.md or skills. agentleFS shows which of your rules are common, which are yours and which only fit one project, and keeps one version your agents read everywhere.

Check your setup, free