القائمة
Music generation. Produces a music clip from a text prompt using Lyria. Use for background scores, mood pieces, jingles, underscore tracks, or any plan node requiring generated audio.
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
Long video production. Composes a video from a sequence of distinct shots. Trigger ONLY when the user explicitly asks to generate a video longer than 10 seconds (duration > 10s). DO NOT trigger if the requested video is 10 seconds or less.
Cinematography vocabulary for shot framing, camera angles, subject relationship, and camera movement. Load when the user's request involves shot-type choices or when constructing ask_user options for framing/movement settings.
Image generation. Produces a still image from text, edits an existing image, or composites multiple images. Use for any output that is a single still frame — concept art, character references, product shots, scene keyframes, style explorations, background plates, virtual try-on, or image editing.
Video generation. Animates a source image into a short video clip, or generates video from text. Use for any output that requires motion — bringing images to life, camera moves, atmospheric animation, or full scene videos.
Implements a feature end-to-end — spec, plan, build, review, simplify, ship — fully autonomous from a feature description file. Use when you have a feature described in a file and want the full pipeline to run without pausing for input.
Character generation and identity board planning. Creates an artistic, high-quality Character Identity Board to establish and lock a character's visual identity (face, clothing, body type, expression range, and silhouette) for future image and video generation.
| name | music-generation |
| description | Music generation. Produces a music clip from a text prompt using Lyria. Use for background scores, mood pieces, jingles, underscore tracks, or any plan node requiring generated audio. |
| metadata | {"type":"primitive"} |
Use t2m when:
Every music prompt must specify four core dimensions. Omitting any dimension lets the model guess, producing generic or mismatched output.
[GENRE & STYLE] [MOOD & EMOTION] [INSTRUMENTATION] [TEMPO & ENERGY]
[SONG STRUCTURE] [PRODUCTION QUALITY]
[VOCALS or "instrumental"]
[NEGATIVE PROMPT (optional)]
Name the genre explicitly. Compound genres are valid and often preferred.
Sub-genre precision improves output — "new-wave synth-pop" is better than "electronic."
Use concrete emotional descriptors rather than vague adjectives.
Pair moods when appropriate: "bittersweet yet hopeful."
Name specific instruments. Generic terms ("guitar," "strings") are weaker than specific ones.
Always specify the lead instrument — it anchors the texture: "lead: solo cello."
Provide BPM when precision matters; use descriptive terms otherwise.
Always pair tempo with an energy word: "slow (40 BPM), sparse and sparse" or "fast (140 BPM), energetic and propulsive."
Optional but powerful for controlling dynamics and arc. Use only when the user has a narrative intent.
For clips (30s default model), one dynamic arc is the maximum — do not plan full song structures.
Sets the sonic aesthetic.
Always declare explicitly.
"instrumental" as a standalone keyword — always include this when vocals are not wanted.vocals singing: "Rise up and take the stage."Use negative_prompt to exclude specific elements. Be precise — model-level exclusions, not abstract moods.
| Want to avoid | Negative prompt value |
|---|---|
| All vocals | "vocals, singing, choir, spoken word" |
| Percussion | "drums, kick, snare, hi-hat, percussion" |
| Distortion | "distorted guitar, fuzz, overdrive, noise" |
| Specific instruments | "violin, piano, brass" |
| Bass frequencies | "bass guitar, sub-bass, 808" |
| Genre bleed | "jazz, hip-hop, electronic" |
Do not use mood words in negative prompts — "sad", "boring", "cheap" have no effect.
For dynamic genre or mood shifts within a clip, use [mm:ss] markers in the prompt. Only available for longer clips (lyria-3-pro-preview).
[0:00] Sparse piano intro, slow and contemplative, 60 BPM.
[0:15] Full orchestral swell enters — strings, brass, timpani.
[0:45] Returns to solo piano, fading out.
Rules:
[mm:ss] format.| Model | Duration | Best for |
|---|---|---|
lyria-3-clip-preview | ~30s | Default. Loops, stingers, short background pieces. |
lyria-3-pro-preview | up to ~3min | Full tracks, complex structures, timestamp prompting. |
Always use lyria-3-clip-preview unless the user explicitly needs a longer track or timestamp-based structure.
Background underscore for a product video:
Upbeat corporate pop. Optimistic and forward-moving. Instrumentation: electric piano, clean electric guitar, light synth pads, punchy kick and snare. 118 BPM. Builds from sparse verse to full chorus at 0:15. Studio quality, polished mix. Instrumental.
Suspense cue for a horror sequence:
Dark cinematic orchestral. Tense, dread-building. Lead: low string ostinato with dissonant brass stabs. Sparse, irregular percussion. 60 BPM rubato. No melody — sustained atmospheric texture. Large-room reverb. Instrumental.
Negative prompt: piano, acoustic guitar, major key, bright tones.
Lo-fi study track:
Lo-fi hip-hop. Melancholic and cozy. Instrumentation: Fender Rhodes, vinyl-sampled double bass, brushed jazz drums, atmospheric pad. 85 BPM. Loopable, no structural variation. Warm tape saturation, vinyl crackle. Instrumental.
Cinematic epic with vocals:
Cinematic epic orchestral. Triumphant and emotionally charged. Full orchestra: soaring strings, French horns, choir, orchestral percussion. 120 BPM, driving. Builds through verse to massive final chorus. Studio quality, IMAX-scale dynamics. Wordless female and male choir.
vocals singing: format — lyrics may appear as narration instead.negative_prompt — only instrument and genre terms have effect.lyria-3-pro-preview and timestamp prompting.