---
name: music
description: "Reference guide for making a song using external tools with Suno V6 through kie.ai — from the person's hum, sung or rapped line (the melody is kept), from an idea, from their own lyrics, in the style of a reference they name (MusicBrainz via music-refs), or as a remake of their track; then stems, a longer version, fixed words, cover art, a music video. The source asked for one creative choice at a time. Points to music-hum, music-write and music-after; the examples use external music tools. Use when the user says song, track, music, beat, hum, rap, cover, stems, трек, песня, музыка, напел, кавер, бит, стемы, минус."
---

These creative references describe source experiments with external tools. Use your own supported account and tools; this download installs no automation. Settings and timings are experimental observations, not current FlashBoards availability. Use only media and voices you have permission to use.

# music · reference workflow

**Work on the person's own voice, words and tracks. The example run below belongs to the owner's test; never reuse its words.**

Slots in all music-* skills: `{seed}` — the person's recording (hum, line, voice message), `{track}` — a finished track,
`{lyrics}`, `{style}`, `{name}` — a short file name for this song.

## 🏆 Source experiment pick — the method

1. **Hear it first.** Any recording → transcription and audio understanding through your supported external tools: the words with timings (Scribe v2) and
   what Gemini 3.8 Flash hears — kind (rap, singing, humming), 3 genres, tempo, delivery, the theme, a song idea. Keep these as project notes; no private analysis JSON format is required.
2. **Resolve one creative choice at a time**: correct uncertain words by listening to the recording, choose the vibe, then decide whether to write new lyrics, use your own or make an instrumental.
   **Use MusicBrainz for the vibe references** (see «Choose a vibe» below): compare named references next to each
   genre, listen to authorized examples and keep a style description based on that data, not from memory. These are ordinary review steps, not chat cards or installed automation.
3. **Write the lyrics and the style** by `music-write`: the person's own line becomes the hook, the rest is new, in the
   language they sang in; the style is one line of genre + voice + instruments + tempo.
4. **Make it** — the external song request: with `source` = `{seed}` (a cover — Suno keeps the melody and the flow), `audio_weight`
   0.65 (closer to the hum 0.75, freer 0.5), `vocal_gender` chosen for the intended singer. **Lyrics are required** for a vocal song: without
   them Suno makes an instrumental (the source workflow refused).
5. **Two takes** land in your supported external workflow with a player → choose: take 1 / take 2 / another round → then what next (`music-after`).

## Classify

| The person has | Route |
|---|---|
| A voice message or a recording of a hum, a sung or rapped line | `music-hum` |
| A reference — "like <artist>", a song or a genre they love | `music-refs` (MusicBrainz: genres, era, similar artists → 3–5 references and the style line) → `music-write` → a new song request without `source` |
| An idea or a theme ("a song about my dog stealing socks") | `music-write` → a new song request without `source` |
| Their own lyrics | `music-write` (style only, keep their words exactly) → a song request |
| A finished track to remake in another style | a song request with `source` = the track (≤ 8 min), lyrics from them or `instrumental: true` |
| A beat or an instrumental for a video | a song request with `instrumental: true`, `duration` = the video length |
| A take they like and want more from | `music-after` |

## Choose a vibe — references from MusicBrainz

Whenever the genre is open (after listening, or an idea with no genre named), compare the following in your project notes:

1. Pick 3 genres: from the recording's sound description, or the 3 that fit the idea and mood best.
2. Look up those genres in MusicBrainz (the source’s data lookup took ~8 s) → for each: the
   MusicBrainz genre («Russian Trap» → trap), 3 recognisable artists and a style seed with no names.
3. List each genre with "like <A>, <B>, <C>" references. Note the best fit, but choose another genre, artist or song if that better matches your intent.
4. The chosen vibe's description starts the style line (`music-write` adds voice, tempo, instruments, mood).
   For an artist or song, use `music-refs`; if MusicBrainz has no matching genre, describe the recording's sound in ordinary words instead.

Artist names are for choosing references only, never in `style`.

## Models and times (9.10.2026, measured)

| Step | Model | Time |
|---|---|---|
| `listen` | Scribe v2 + Gemini 3.8 Flash | ~15 s |
| `vibes` / `find` / `style` | MusicBrainz (source experiment’s prepared index, not bundled) | 3–8 s |
| `make` (2 takes, up to 6 min) | Suno V6 | 25–80 s |
| `stems` vocal + instrumental | Suno V6 | ~1 min |
| `extend` (2 takes) | Suno V6 | ~1 min |

The table preserves the source's measured batch. A separate same-day chat prototype took 121 s for two 2:29 takes (`music-hum`); the 25–80 s range is not a universal completion time. Step names here describe external tasks, not bundled commands.

Choose the settings and confirm the request through your supported external tool before generating.

## Global rules

1. **Lyrics in the language of the seed** (a Russian rap stays Russian), even though you talk to the person in English.
2. **Never write out the words of an existing song** from memory. For a remake with the original words, the person pastes them.
3. **Name no artist in `style`** — describe the sound (genre, voice, instruments, tempo); Suno blocks artist names. When the person names a reference, the genres, era and similar artists come from MusicBrainz (`music-refs`), not from memory.
4. **Their line stays theirs**: keep the words of the seed as sung (fix only what they correct), build the hook from them.
5. Save both returned takes and their original task/track identifiers with your project notes.
6. If a request is interrupted, check that same task through your supported tool before submitting another request.
7. **A cover needs the person's own audio.** Suno refuses a commercial recording as `source` (kie code 413 «matches an existing recording in our catalog», 2 of 2 on 9.10.2026) — offer a new song in that style; never alter a recording to get past the check.

## Example run (reference only — never reuse its words or files)

9.10.2026, the owner's 8 s voice message (Russian rap, male, ~125–135 BPM). `listen`: Scribe and Gemini agreed on 3 of 4
lines and split on one word (shown as a card in the source prototype, not a required reader interface). Hook from the seed, two verses and a bridge written around «flying above the
ground», style «russian cloud trap, male rap vocal in Russian, laid-back hypnotic flow, ad-libs, 125 BPM, dreamy pads,
gliding 808, rolling hi-hats», `audio_weight` 0.65, `duration` 150 → two 2:30 takes in 63 s. Without lyrics the
same source gave two instrumentals — that is why the source workflow refused a vocal song without words.