AI Music & Lyrics Prompt Builder

Structure is what turns an idea into a usable music prompt

When you create with an AI music tool — 天工 SkyMusic, 海绵音乐 (Hisound) or Suno — the quality of the track depends far more on how clearly you describe what you want than on any single magic word. A good music prompt names the genre, sets the mood, fixes the tempo, lists the instruments, chooses the vocals, picks the language, and supplies a lyric theme or the lyrics themselves. This builder keeps that structure for you: fill the fields and it joins them into a clean prompt with a labelled style block — genre, mood, tempo, instruments, vocals and language each on their own line — followed by an optional lyrics block, ready to paste into any AI music tool. The result is the kind of brief a careful music producer would write by hand, only assembled in seconds and entirely in your browser.

The two highest-leverage fields are genre and mood, because they are the first things an AI music tool latches onto. "Warm acoustic folk, nostalgic" steers the whole arrangement in one short line, far more efficiently than a paragraph of adjectives. After that, tempo and instruments do the shaping: tempo decides whether the song breathes or drives, and the instrument list — "piano and strings", "808 drums and synth bass" — tells the model what the track should be built from. You do not need a music-theory vocabulary to do this well; plain, concrete words ("slow", "upbeat", "acoustic guitar", "lo-fi") are exactly what these tools are trained to read, and they travel across Mandarin, Cantonese and English prompts alike.

Vocals, language and lyrics decide whether the result is your song

The fields people skip and regret are vocals, language and lyrics. The vocal choice — male, female, choir, or instrumental only — changes the entire feel of the result, and stating it removes guesswork. Language matters too: Mandarin, Cantonese and English carry different diction and rhythm, so naming the sung language makes the generated vocal sound natural rather than forced. And lyrics are where a generic demo becomes your song: you can leave the box blank and let the AI write from a theme, or paste your own finished words, which the builder keeps in a separate, clearly marked lyrics block so the tool reads them as lyrics rather than as style notes. Spending one extra line on vocals and language is consistently the cheapest way to make the output sound intentional.

Because the output is structured plain text, the same prompt is portable across every major AI music tool — what works in 天工 carries over to 海绵音乐 and Suno with little change, since each reads the same kind of descriptive brief. That makes iterating cheap: tweak one field, copy again, and re-generate. Treat the first prompt as a draft. Run it, listen to where the track drifts from what you imagined, and tighten the matching field — slow the tempo, swap the lead instrument, sharpen the mood word. Two or three rounds of that usually turn a passable demo into something close to what you heard in your head, and you keep a clean, reusable prompt at the end. And because the whole tool runs locally in your browser, you can experiment freely without anything you type ever leaving your device, being sent to a model, or being stored. This tool has no official affiliation with 天工, 海绵音乐 or Suno — it simply helps you write a better brief for whichever one you use.