القائمة
Generate music using Gemini Lyria. Use for background music, jingles, intros, soundtracks, and complete songs WITH sung vocals — Lyria 3 sings your exact lyrics (8 languages, rap/sung delivery).
التثبيت باستخدام Codex أو Claude انسخ هذا Prompt والصقه في Codex أو Claude أو مساعد آخر ليراجع صفحة Skill ويثبّتها لك.
استنادا إلى تصنيف SOC المهني
MANDATORY pre-flight check that runs BEFORE any content plan or generation. Classifies the topic, verifies project/brand/product context, resolves the required reusable assets (characters, products, logos, voices) from the asset registry, and provides/generates whatever is missing — saving paths for reuse. Use whenever the user asks to plan a campaign, make a marketing plan, or create/generate any image, video, voiceover, or music.
Generate images using Gemini 3 Image (Nano Banana). Use for thumbnails, carousels, infographics, social graphics, product photos.
Generate video content using Gemini Veo. Use when user asks to create video, TikTok ad, Instagram Reel, YouTube video, product video, explainer.
Generate voiceover audio using Gemini TTS. Use for narration, voiceovers, podcasts, multi-speaker dialogue.
Write the words — scroll-stopping hooks, framework-driven scripts, and platform-ready captions with hashtags. Use when the user needs hook ideas, a video/ad script, ad copy, post captions, or says "what should I post", "write a script", "give me hooks", "write the caption".
Analyze a reference video (YouTube URL or local file) into a scene-by-scene breakdown — timecodes, shot types, camera moves, on-screen text, spoken lines — plus a recreation blueprint with per-scene video prompts and VO script. Use when the user says "analyze this video", "I want to create something like that", "recreate this video", "break down this YouTube video", or pastes a video link they want to emulate.
| name | generate-music |
| description | Generate music using Gemini Lyria. Use for background music, jingles, intros, soundtracks, and complete songs WITH sung vocals — Lyria 3 sings your exact lyrics (8 languages, rap/sung delivery). |
| allowed-tools | Bash Read Write Edit |
Do NOT create a projects/{name}/scripts/*.ts file. Run the workflow command:
node workflows/cli.cjs generateMusicTrack \
'{"prompt":"...","outputPath":"projects/{name}/output-contents/track.wav","duration":30}'
The CLI loads .env for you and prints the JSON result (path + cost). Other commands
you may chain: mixVideoAudio, and the asset helpers loadAssetConfig / resolveAsset /
registerAsset (pass positional args as a JSON array, e.g. loadAssetConfig '["myproject"]').
Argument shapes are in workflows/WORKFLOWS.md.
Output Location: projects/{name}/output-contents/ (pass this as outputPath in the args)
Does projects/{project-name}/ exist?
YES → Read these files for context:
NO → Ask user directly
First, check the registry for a reusable music bed — don't regenerate the same track:
node workflows/cli.cjs loadAssetConfig '["{name}"]'
node workflows/cli.cjs resolveAsset '["{name}","music-main"]' # existing project music bed? → { "existing": [ ... ] }
bed.existing[0]) instead of generating. Confirm with user.If project exists, READ and USE:
From projects/{name}/templates/project.md:
product.name → Context for what the music supportsaudience.primary.age_range → Age-appropriate music styleFrom projects/{name}/templates/brand.md:
tone.energy → High/medium/calm → determines tempotone.voice → Professional/casual → determines genrevisual.style → Modern/classic → influences music style"What is this music for?"
Genre?
Mood?
Duration?
Tempo?
Vocals or instrumental? (Lyria 3 generates vocals BY DEFAULT)
| Model | Duration | Cost | Use For |
|---|---|---|---|
lyria-3 | ≤30 seconds | $0.04 | Intros, outros, short clips, jingles |
lyria-3-pro | ≤3 minutes | $0.08 | Full songs (incl. sung vocals + your lyrics), longer background music |
Lyria 3 generates vocals and lyrics BY DEFAULT. It can sing your EXACT lyrics,
including rap delivery, backing vocals, and structured arrangements. Everything goes in
the prompt string — there is no separate lyrics parameter.
How to provide exact lyrics — prefix the lines with Lyrics: inside the prompt:
...style/instrumentation description... A confident male vocalist performs the verses
with a rhythmic rapping delivery. Structure: intro; verse one; INSTRUMENTAL chorus;
verse two; outro spoken line. Only the lines below are sung.
Lyrics (verse one, rapped):
<exact lines>
Lyrics (verse two, rapped):
<exact lines>
Steer the voice: gender, timbre/range ("smooth baritone", "breathy soprano"), delivery ("rapping", "spoken"), and language. Mention where backing vocals should echo the lead if wanted.
Supported vocal languages (8): English, German, Spanish, French, Hindi, Japanese,
Korean, Portuguese. Burmese/Myanmar is NOT officially supported — output is a
retry lottery: one production run sang near-real Burmese (transcribed back as actual
Burmese script, close to the target lyrics), another gave pure vocalese/gibberish.
If the user wants MM vocals, attempts are cheap ($0.08) — generate, verify with
transcribeAudio + a native listen, retry if garbled. The guaranteed route is still:
generate MM sections INSTRUMENTAL (state it explicitly per section: "a fuller
INSTRUMENTAL chorus with no vocals where piano and strings carry the melody") and
overlay a real recording later.
Instrumental-only tracks: because vocals are the default, ALWAYS end background-music prompts with "Fully instrumental, no vocals."
Production-tested gotchas (cost real money if missed):
quality:"standard" may 404 on some keys — use quality:"pro" (auto-falls back
to lyria-3-pro-preview, same price).ffmpeg -f mp3 -i out.wav -c:a pcm_s16le fixed.wavtranscribeAudio on the result (~$0.01) and diff
against the locked lyrics before calling it done.Combine user preferences with brand context:
[Genre matching brand.tone.voice] music with [instruments],
[mood matching brand.tone.energy] and [tempo] tempo,
[purpose/context],
[any specific elements]
Map brand.tone to music:
brand.tone.energy: "high" → Fast tempo, energetic, upbeatbrand.tone.energy: "medium" → Medium tempo, confidentbrand.tone.energy: "calm" → Slow tempo, relaxingbrand.tone.voice: "professional" → Corporate, clean soundbrand.tone.voice: "casual" → Pop, acoustic, friendlybrand.tone.voice: "bold" → Cinematic, dramaticExamples:
For tech product video (energetic brand):
Upbeat electronic music with synth leads and punchy drums,
energetic and modern, medium-fast tempo,
suitable for tech product showcase video,
building energy with drops
For wellness brand (calm brand):
Calm ambient music with soft piano and gentle synth pads,
relaxing and peaceful, slow tempo,
for wellness brand background,
minimal and spacious sound
For corporate/professional brand:
Professional corporate music with light acoustic guitar,
confident and optimistic, medium tempo,
for business presentation background,
subtle and non-distracting
Propose every choice with reasoning from brand.md, then show one summary and wait for go —
let the user change any field. Don't auto-generate.
🎵 MUSIC SETUP — confirm or adjust
Purpose: background for product video
Genre: upbeat electronic (suggested — matches your "energetic" brand)
Mood: modern, confident
Tempo: medium-fast
Length: 30s → quality: standard (lyria-3, ~$0.04)
Prompt: "Upbeat electronic music with synth leads…"
Output: projects/{name}/output-contents/music.wav
Proceed, or change a field? (proceed / slower / longer / different genre …)
If the user says "just do it", still show this summary once and state the assumed defaults.
Only after confirmation. Use the pre-built workflow function — do NOT re-implement file
saving, retry, or cost handling. It validates input, enforces the duration limit per quality,
retries on transient errors, writes the .wav, and returns the cost.
# Short clip (standard = lyria-3, ≤30s) - ~$0.04
node workflows/cli.cjs generateMusicTrack '{"prompt":"Your music description here","outputPath":"projects/{name}/output-contents/music.wav","quality":"standard","durationSeconds":30}'
# → { "success": true, "data": { "audioPath": "...", "cost": { "totalCost": 0.04 } } }
# On failure → { "success": false, "error": { "code": "...", "message": "..." } }
# Full song (pro = lyria-3-pro, ≤180s) - ~$0.08
node workflows/cli.cjs generateMusicTrack '{"prompt":"Your music description here","outputPath":"projects/{name}/output-contents/song.wav","quality":"pro","durationSeconds":120}'
# On success, save the bed to the registry so it can be reused (ask if it should be the project bed)
node workflows/cli.cjs registerAsset '["{name}","music",{"id":"music-main","label":"Brand music bed","description":"Upbeat electronic, energetic, medium-fast","source":"generated","path":"assets/music/music-main.wav","duration_seconds":30},{"date":"<today ISO>"}]' # copy the output .wav to path for reuse
registerAssetis its own CLI command (positional JSON array args), same asgenerateMusicTrack.
Use the mixVideoAudio workflow — it handles the ffmpeg mix (music ducked under a
voiceover if one is provided) and returns a clean result:
node workflows/cli.cjs mixVideoAudio '{"videoPath":"projects/{name}/output-contents/video.mp4","musicPath":"projects/{name}/output-contents/music.wav","musicVolume":0.3,"outputPath":"projects/{name}/output-contents/final.mp4"}'
# omit voiceoverPath, or add "voiceoverPath":"..." to layer both (music auto-ducks); musicVolume is 0–1
For a full deliverable (multiple clips + VO + music + captions in one call) use
assembleFinal() — see workflows/WORKFLOWS.md.
Save to: projects/{name}/output-contents/ or current directory
music.wavmusic-prompt.txt (keep the prompt for reference)<content-id>.pipeline.json IN the content folder BEFORE generating
(nodes = CLI commands, {{node.data.field}} refs), then
node workflows/cli.cjs runPipeline @<file> — see workflows/pipelines/README.md.prompts.txt with the music prompt; update it on
every retry with a one-line RESULT note.createGenerationManifest / addManifestEntry
CLI commands (AGENT-GUIDE Step 5).Lyrics: blocks (Step 3b)