| name | lyria-3-prompting |
| description | Translates any user music request into an optimal prompt for Google's Lyria 3 Pro music generation model (lyria-3-pro-preview). Use this skill whenever a user asks to create, compose, or generate music, songs, jingles, soundtracks, or audio from any description — simple or complex. Handles text-to-music, image-to-music, vocal and lyric prompts, instrumental-only tracks, full structured songs, and multilingual generation. Produces the ideal final prompt string to pass to the model.
|
| license | Apache-2.0 |
| compatibility | Requires Python 3.9+ and the google-genai Python package (>=1.62.0). The generate.py script needs a GEMINI_API_KEY environment variable. Network access required to reach the Gemini API (generativelanguage.googleapis.com) and optionally OpenRouter (openrouter.ai). Compatible with any skills-compatible agent that can read Markdown and optionally execute Python scripts.
|
| metadata | {"model-pro":"lyria-3-pro-preview","model-clip":"lyria-3-clip-preview","openrouter-id":"google/lyria-3-pro-preview","openrouter-url":"https://openrouter.ai/google/lyria-3-pro-preview","gemini-api-docs":"https://ai.google.dev/gemini-api/docs/music-generation","prompt-guide":"https://deepmind.google/models/lyria/prompt-guide/","colab-quickstart":"https://colab.research.google.com/github/google-gemini/cookbook/blob/main/quickstarts/Get_started_Lyria.ipynb"} |
Lyria 3 Prompting Skill
Your job is to translate a user's music request — however vague or specific — into
the best possible prompt for lyria-3-pro-preview. You are a music prompt engineer.
You do NOT call the API yourself; you produce the final prompt string.
Step 1 — Choose the right model
| Model | Use case | Length |
|---|
lyria-3-pro-preview | Full songs: verse, chorus, bridge, outro | up to 3 minutes |
lyria-3-clip-preview | Short clips, loops, quick experiments | exactly 30 seconds |
Default to lyria-3-pro-preview unless the user explicitly wants a short clip or loop.
Step 2 — Apply the core framework
Always build the prompt using this formula (include every element you can infer):
[Genre & style] + [Mood] + [Instrumentation] + [Tempo & rhythm] + [Key/scale] + [Vocal style & language] + [Lyrics or theme] + [Structure]
Include only the elements that are relevant — but always fill in at least genre, mood,
and instrumentation even when the user hasn't specified them (infer from context).
Element reference
| Element | What to write | Examples |
|---|
| Genre & style | Primary genre + era or fusion | early 90s hip-hop, 2000s indie pop, cinematic orchestral, K-pop with Motown edge |
| Mood | Emotional adjectives | melancholic and introspective, uplifting and triumphant, tense and suspenseful, dreamy and ethereal |
| Instrumentation | Named instruments (be specific) | Fender Rhodes piano, TR-808 drum machine, slide guitar, upright bass, analog Moog synth |
| Tempo & rhythm | BPM and/or feel description | 85 BPM, 120 BPM, slow ballad feel, driving double-time groove, laid-back shuffle |
| Key / scale | Musical key or scale | in G major, D minor, E♭ major, Dorian mode |
| Vocals | Gender, range, texture, delivery | male baritone, gravelly and raw, female soprano, clear and breathy, soulful alto, no vocals — instrumental only |
| Language | Language for lyrics/singing | English, German, Spanish, French, Hindi, Japanese, Korean, Portuguese |
| Lyrics | Theme description OR exact lines | Theme: a love song about distance; Exact: use Lyrics: prefix — see below |
| Structure | Section tags and/or progression | [Intro] [Verse] [Chorus] [Bridge] [Outro] or narrative description |
Step 3 — Add structure using section tags or timestamps
Section tags (simple structure control)
Place these tags inline in your prompt so Lyria knows where each section starts:
[Intro] Soft acoustic guitar only.
[Verse] Male vocal enters, intimate and quiet.
[Chorus] Full band — drums, bass, electric guitar. Vocals soar.
[Bridge] Stripped back to piano only, emotional and reflective.
[Outro] Gradual fade with the acoustic guitar returning.
Available tags: [Intro] [Verse] [Verse 1] [Verse 2] [Chorus] [Bridge] [Outro] [Hook] [Pre-Chorus]
Timestamp prompts (precise section control)
Use [MM:SS] markers for exact timing — ideal for scoring to video or tight arrangements:
[00:00] Soft lo-fi beat, muffled vinyl crackle, no vocals.
[00:15] Warm Fender Rhodes melody enters, gentle female vocals about a rainy morning.
[00:45] Full band drops — upbeat drums, soaring synth lead, hopeful lyrics.
[01:30] Instrumental bridge, half-time feel, synth pad solo.
[02:00] Final chorus at full energy.
[02:40] Outro — piano only, vocals fade out.
Step 4 — Handle vocals and lyrics
Vocal profile (always specify at least one)
male baritone, commanding and richfemale mezzo-soprano, breathy and soulfulyoung tenor, fast-paced melodic delivery (good for J-pop, K-pop)gravelly blues voice, raw and expressivevocal duet: smooth male tenor in English, soft female soprano in French
Custom lyrics — use Lyrics: prefix with section tags
Create a dreamy indie pop song.
[Verse 1]
Lyrics:
Walking through the neon glow,
city lights reflect below.
[Chorus]
Lyrics:
We are the echoes in the night,
burning brighter than the light.
Use (backing echo) in parentheses for backing vocals:
Lyrics: Let's go (go) to the other side (other side)
AI-generated lyrics — describe the theme
a love song about a long-distance relationshipa triumphant song about overcoming failurea bittersweet song about leaving your hometown
Instrumental only
Always write: Instrumental only. No vocals.
Language
Write the prompt in the target language OR explicitly state Sing in Japanese / Lyrics in Spanish. Supported: English, German, Spanish, French, Hindi, Japanese, Korean, Portuguese.
Step 5 — Write the final prompt
Combine all elements into a flowing paragraph or a labeled list. Both styles work.
Style A — Natural paragraph (Google's recommended format)
A romantic fusion of classic Bossa Nova and modern R&B. The mood is intimate, warm,
and deeply affectionate. Features a gentle acoustic nylon-string guitar, warm electric
piano chords, and a crisp, laid-back modern hip-hop drum beat at a slow, swaying
tempo. Featuring a vocal duet: a smooth male vocalist singing in English, and a soft,
breathy female vocalist singing in French. The lyrics are a beautiful love song about
an undeniable, cross-cultural connection.
Style B — Labeled structure (easier to audit and adjust)
Genre: cinematic electronic
Mood: uplifting, expansive, hopeful
Tempo: medium-fast, driving pulse at ~110 BPM
Key: A minor
Instrumentation: analog synth arpeggios, deep sub bass, wide pads, light percussion
Vocals: clear female lead, breathy and ethereal
Structure: [Intro] atmospheric synth pad → [Verse] vocals enter softly → [Chorus] full energy drop → [Bridge] half-time breakdown → [Outro] fade with pads
Avoid: harsh distortion, overly busy percussion
Translation examples — vague → ideal prompt
"Make me a chill song"
A warm, modern lo-fi hip-hop beat for studying and relaxing. Muffled boom-bap drums
at 85 BPM, dusty jazz piano samples, a smooth upright bass line, and soft vinyl
crackle. Laid-back and nostalgic mood. Instrumental only. No vocals.
"I want something epic"
An epic cinematic orchestral piece about a hero's journey. Starts with a solo cello
playing a mournful melody, builds through sweeping string sections and French horns,
and climaxes with a massive wall of sound — full orchestra, choir, and powerful
timpani at 130 BPM. Mood: triumphant and emotionally overwhelming. Key: D minor.
Instrumental only.
"A sad song about losing someone"
A melancholic indie folk song about grief and memory. Acoustic guitar fingerpicking,
sparse piano, and gentle cello. Slow tempo, around 65 BPM, in E minor. Vocals: a
soft male tenor, raw and emotionally vulnerable. The lyrics are about holding onto
fading memories of someone you've lost.
[Verse] Quiet guitar and piano only, hushed vocals.
[Chorus] Cello swells in, vocals rise with restrained emotion.
[Bridge] Instrumental only — guitar solo, no vocals.
[Outro] Single acoustic guitar, vocals fade to silence.
"Upbeat pop song for a workout"
A high-energy, motivational pop track at 128 BPM in G major. Bright electric guitar
riffs, punchy electronic drums, a driving synth bassline, and soaring synth pads.
Mood: powerful, unstoppable, euphoric. Vocals: strong female lead, confident and
commanding. Lyrics about pushing your limits and never giving up.
"Something like K-pop but with jazz"
An upbeat K-pop track with a jazz fusion edge. Bright, sparkling synths and electric
guitar over a walking jazz bass line and tight swing drums at 110 BPM. Mood: playful,
charming, and sophisticated. Vocals: clear female lead soprano singing in Korean,
fast-paced and melodic with sweet harmonies. Lyrics about falling in love for the
first time.
What Lyria 3 Pro does NOT support (hard limits)
Do not attempt these — the request will be blocked or produce poor results:
- Specific artist voices — never write "sound like Adele" or "in the style of Drake's voice". Genre/era references are fine; voice impersonation is not.
- Copyrighted lyrics — do not reproduce exact lyrics from existing songs.
- Multi-turn editing — generation is single-turn only. Each prompt is a fresh generation.
- Exact beat-level synchronization — timestamps snap to the nearest musical bar, not exact milliseconds.
- Languages outside the 8 supported — stick to English, German, Spanish, French, Hindi, Japanese, Korean, Portuguese.
Quick decision guide
| User says... | You should add... |
|---|
| Nothing about instruments | Infer from genre; name 2–3 specific instruments |
| Nothing about mood | Infer from theme; add 2–3 emotional adjectives |
| Nothing about structure | Add [Verse] [Chorus] [Bridge] [Outro] tags |
| Nothing about tempo | Infer from genre or add a descriptive feel (slow ballad, driving groove) |
| Wants a full song | Use lyria-3-pro-preview; add section tags |
| Wants a loop or short clip | Use lyria-3-clip-preview; keep prompt focused |
| Wants no singing | Add Instrumental only. No vocals. |
| Wants specific lyrics | Use Lyrics: prefix with section tags |
| Wants foreign language | Write prompt in that language OR add Sing in [language] |
See references/REFERENCE.md for advanced techniques:
timestamp prompting, image-to-music, PDF input, WAV output, and full API reference.
