# Music Video Production > Provider-independent production workflow for AI agents creating music videos, lyric videos, visualizers, performance/narrative/dance videos, social cutdowns, album or single launch assets, and generated-media clips synchronized to songs. Use when planning, directing, prompting, editing, reviewing, or delivering video content built around a recorded song, artist performance, lyrics, beat timing, c… - **Type:** Skill - **Install:** `agentstack add skill-calesthio-generative-media-skills-music-video-production` - **Verified:** Yes — security-reviewed for prompt injection and unsafe behavior - **Seller:** [calesthio](https://agentstack.voostack.com/s/calesthio) - **Installs:** 0 - **Category:** [Developer Tools](https://agentstack.voostack.com/c/developer-tools) - **Latest version:** 0.1.0 - **License:** MIT - **Upstream author:** [calesthio](https://github.com/calesthio) - **Source:** https://github.com/calesthio/generative-media-skills/tree/main/skills/production/content-formats/music-video-production - **Website:** https://github.com/calesthio/OpenMontage ## Install ```sh agentstack add skill-calesthio-generative-media-skills-music-video-production ``` Requires the [AgentStack CLI](https://agentstack.voostack.com/docs/cli). Works with Claude Code, Cursor, and any MCP-compatible agent. ## About # Music Video Production Use this skill to turn a song into a coherent, rights-aware, platform-ready video package. Treat the song as the fixed spine: the video exists to amplify its structure, artist identity, lyrics, and release goal. Do not treat this as generic video generation. Music videos need beat maps, phrase maps, lyric accuracy, sync review, rights notes, performance direction, platform-specific cutdowns, and approval checkpoints. ## Ground rules - Keep the master audio unchanged unless the artist, label, or client approves a radio edit, clean edit, stem edit, speed change, fade, or shortened version. - Never invent rights clearance. Record what is cleared, unknown, platform-provided, licensed, royalty-free, public-domain, or generated. - Do not provide legal advice. Flag clearance, likeness, endorsement, union, label, distributor, or platform-risk questions for the client, rights holder, label, manager, or counsel. - Re-check platform rules at production time. Music libraries, AI labels, commercial disclosure tools, ad specs, and upload limits change often. - Separate documented facts from production heuristics. Facts should trace to current sources; heuristics are creative recommendations. - Preserve provenance: song source, lyric source, stems, edits, models, prompts, seeds when available, consent documents, licenses, platform checks, and review notes. ## Intake Collect enough information before concepting: 1. Song package - final audio file, duration, sample rate, bit depth, loudness target if known - song title, artist name, featured artists, version name, ISRC/UPC if released - approved lyrics, explicit/clean status, lyric translation if needed - optional stems: vocal, instrumental, drums, bass, melody, ad-libs - BPM, key, time signature, section labels if known 2. Release and audience - goal: premiere, visualizer, lyric video, fan engagement, paid campaign, tour promo, album rollout, short-form hook testing - target platforms and aspect ratios - due date, review approvers, approval order, locked brand/artist guidelines 3. Rights and safety - who owns or controls the master and composition - whether sync use, social use, paid ad use, territory, term, and platform rights are cleared - whether any samples, interpolations, remixes, covers, or third-party stems are present - whether any brand, product, venue, artwork, tattoo, logo, prop, or recognizable person needs clearance - whether AI-generated video, voice, likeness, choreography references, or style references will be used 4. Creative references - 2-5 reference videos with what to borrow: pacing, lighting, blocking, edit rhythm, typographic treatment, camera language, color, surrealism level - what not to copy: artist likenesses, exact choreography, protected logos, distinctive set pieces, shot-for-shot structures If rights status is unknown, proceed only with planning or internal mockups. Do not recommend publishing, paid promotion, distribution delivery, or public release until clearance is resolved. ## Choose the concept route Pick one primary route and one secondary texture. A video can blend modes, but it should not feel like six unrelated concepts. - Performance video: artist or performer carries the emotional center. Best when charisma, vocal delivery, band energy, or star identity matters. - Narrative video: story scenes dramatize the song's emotional arc. Best when lyrics imply conflict, transformation, memory, romance, rebellion, grief, or triumph. - Lyric video: typography, timing, and design carry the experience. Best before the official video, when lyrics are a key hook, or when budget/time is tight. - Visualizer: repeated motif, loop, ambient world, generative motion, or symbolic animation supports listening. Best for albums, streaming canvases, and low-friction rollout assets. - Dance or choreography video: movement is the hook. Best for rhythmic tracks, short-form challenges, fan participation, and performance-led social campaigns. - Abstract/generated-media video: AI clips, animation, collage, simulations, or surreal imagery create a dream logic around the song. Best when the artist world is more important than literal story. - Launch asset suite: combine hero video, vertical hook cuts, teaser loops, Canvas-style loops, lyric snippets, behind-the-scenes, and platform thumbnails. Production heuristic: if the song has a memorable chorus lyric, design at least one repeatable visual signature for that phrase. If the song has an instrumental drop, design a non-lyric visual payoff: dance hit, camera move, color switch, surreal transformation, or montage burst. ## Build the song map Create a beat-and-phrase map before writing shot lists or prompts. Minimum map: - timecode in `mm:ss.mmm` - section: intro, verse, pre-chorus, chorus, post-chorus, bridge, breakdown, outro - bar or phrase count if available - key lyric fragments and vocal entrances - beat accents, fills, drops, stops, pickup notes, ad-libs - emotional intensity from 1-5 - edit density target: long holds, medium cuts, rapid cuts, strobe risk, silence - visual event: reveal, movement change, text entry, performance close-up, transition, story beat Example beat map row: ```text 00:45.200-01:07.900 | Chorus 1 | hook repeats twice | intensity 5 | cut on snare every 2 bars; hold final hook for 2.5s | lyric text enters on vocal, not before | hero blue-to-red lighting shift ``` Phrase mapping matters more than raw BPM. Cut on musical reasons: downbeats, lyric starts, drum fills, breath gaps, rhyme completions, bass drops, harmonic changes, and silence. Avoid cutting mechanically on every beat unless the genre and concept reward that intensity. ## Script, treatment, and shot list Write a short treatment before prompts or filming. Treatment should include: - one-sentence promise: what the viewer will remember - artist portrayal: iconic, vulnerable, playful, untouchable, documentary, mythic, anonymous, animated - world: location, era, color, texture, weather, lighting, lens feel - structure by song section - recurring motifs and how they evolve - what appears in chorus 1, chorus 2, final chorus, and why each escalates - rules for lyrics/on-screen text - rules for generated clips: realism level, camera language, continuity anchors, negative constraints - approval risks: rights, likeness, flashing, explicit content, brand safety Shot list fields: ```text ID | Timecode | Section | Shot size | Subject | Action | Camera/motion | Lighting/color | Sync requirement | Asset/source | Notes ``` Sync requirement examples: - `lip-sync exact to lead vocal` - `dance hit on downbeat` - `lyric word appears on syllable` - `cutaway only; no mouth movement` - `loop must restart invisibly every 4 bars` - `AI clip can drift; use under instrumental texture` ## Performance and choreography direction For performance: - Use playback on set or in generated/lip-sync direction. The performer must hear the final approved audio version or a timecode-aligned playback version. - Direct emotional beats by section, not just "sing with energy." Example: verse 1 restrained eye contact, pre-chorus pacing, chorus direct-to-lens, bridge turned away, final chorus full-body release. - Capture sync-safe coverage: locked close-up of mouth, medium performance, wide body, detail inserts, overcranked movement, non-singing cutaways. - Avoid visible singing in shots not intended for exact sync unless the mouth is obscured or off-axis. - For bands, map featured instruments to audible moments. Do not cut to a drummer during a guitar-only fill unless intentionally ironic. For dance: - Mark counts against the song map. Note the move name, body direction, camera direction, and hit frame. - Keep a clean master wide for choreography continuity, then add close-ups and detail shots. - Plan short-form challenge sections around 5-15 seconds of memorable movement. - If AI-generated dancers are used, prompt for broad rhythm and pose change but verify manually; generated motion may not obey exact counts. ## Lyric and on-screen text Use approved lyrics only. If lyrics are not provided, ask for them or transcribe and mark as unapproved. Lyric video rules: - Align words to the lead vocal, not to the instrumental grid. - Keep line breaks musical: phrase, rhyme, breath, or semantic unit. - Design hierarchy for lead vocal, background vocal, ad-lib, and non-lyric captions. - Do not obscure the artist's face at emotional moments unless the concept is typographic-first. - Preserve safe margins for vertical UI overlays and captions. - Provide a clean textless master when feasible for localization and future edits. On-screen text can include title cards, credits, warnings, release date, merch/tour CTA, and disclosures. Keep disclosures readable and near the claim or content they clarify. ## Generated-video prompt translation When using AI image/video tools, translate music-video direction into production prompts. Do not ask a model to "make a music video" and hope for sync. Most generated-video systems are weak at exact lip sync, lyric timing, text fidelity, and continuity unless the provider specifically supports those controls. Prompt components: - shot duration and aspect ratio - subject identity or fictional character description; avoid unauthorized real-person likeness - action in one clear motion phrase - camera movement and shot size - lighting, lens, color, texture, location - timing anchor from the song map - continuity anchors: wardrobe, prop, color, environment - negative constraints: no readable random text, no extra fingers, no logo, no mouth singing, no face morphing, no flicker Example generated-clip prompt: ```text 4-second vertical 9:16 shot for the final chorus of an indie-pop music video. A fictional silver-jacketed performer runs through a rain-slick neon alley, turns sharply toward camera on the downbeat, and throws a handful of glitter that becomes floating stars. Handheld tracking shot, 35mm lens feel, saturated cyan and magenta, wet pavement reflections, high-energy chorus payoff. No real celebrity likeness, no readable signs, no lip-sync mouth movement, no strobing, no brand logos. ``` Use generated clips where they are strongest: surreal inserts, visualizers, environments, transitions, symbolic action, dream sequences, and texture. Use specialist lip-sync or avatar tools, live footage, or carefully directed animation when mouth sync is mission-critical. ## Edit rhythm Music-video editing should feel musically intentional without becoming predictable. - Establish a visual grammar in the intro. - Use verses for story, intimacy, or world-building. - Use pre-choruses for acceleration: camera movement, shorter cuts, rising text density, color transition. - Use choruses for the most repeatable visual idea. - Save one escalation for the final chorus: new location, wider shot, choreography group, color inversion, rain/fire/confetti, crowd, or abstract transformation. - Let important lyrics breathe. A powerful line often lands better with a hold than with more cuts. - Create contrast: dense montage after a long hold, silence after chaos, close-up after wide choreography. - Avoid flash patterns above accessibility thresholds; do not use rapid full-frame strobing as a default "energy" solution. Use cutdowns intentionally: - 6-8 seconds: hook/teaser/Canvas-like loop; one idea only. - 10-15 seconds: social challenge or chorus hook. - 20-30 seconds: paid social or release announcement. - 45-60 seconds: extended short-form excerpt. - Full-length: canonical music video or visualizer. ## Lip-sync and playback sync For live-action or avatar/lip-sync work: - Lock the audio version before generating or filming sync shots. - Use slate/timecode or visible clap for live shoots where possible. - Align mouth shapes to consonants and phrase starts; vowels alone are forgiving, plosives are not. - Check sync at normal speed and half speed. - If a shot is slightly off, use cutaways over the worst syllables instead of stretching the master audio. - Do not alter pitch, tempo, or timing of the released song to fit generated visuals unless explicitly approved. For lyric sync: - Spot check all chorus repeats; copy-pasted timing often drifts because vocal phrasing changes. - Confirm censored/clean lyric treatment matches the distributed audio version. ## Rights, credits, and disclosure checkpoints Escalate to the client, rights holder, label, manager, distributor, platform specialist, or counsel when any of these appear: - copyrighted song is not owned/controlled by the artist or client - cover, remix, sample, interpolation, beat lease, stock loop, or third-party vocal is present - the video will be used in paid ads, brand partnerships, political contexts, regulated industries, or endorsements - real person likeness, voice clone, digital replica, archival footage, fan footage, minors, private locations, tattoos, logos, or artwork appear - platform-native music library is being used outside its allowed context - generated media imitates a living artist, celebrity, dancer, photographer, director, or recognizable visual style too closely - union performer, label contract, distributor, or venue rules may apply Documented facts to apply, verified July 11, 2026: - YouTube requires creators to disclose realistic altered or synthetic content that viewers could mistake for a real person, place, scene, or event; YouTube says clearly unrealistic, animated, special-effects, or production-assistance uses are not generally required under that tool. Source: YouTube Blog, "How we're helping creators disclose altered or synthetic content" and YouTube Help "How this content was made" disclosure pages. - TikTok says creators must label realistic AI-generated content and offers creator and automatic AI labels; misleadingly labeling unaltered content as AI-generated may violate its terms. Source: TikTok Help, "About AI-generated content." - Meta's misinformation policy states that organic content with photorealistic video or realistic-sounding audio must be disclosed using its AI-disclosure tool. Source: Meta Transparency Center, "Misinformation." - FTC endorsement guidance requires clear and conspicuous disclosure of unexpected material connections between an endorser and advertiser. Source: FTC Endorsement Guides and 16 CFR Part 255. - TikTok says business/promotional content should use its Commercial Music Library because music outside that library is not licensed for commercial use in content. Source: TikTok Help, "Commercial use of music on TikTok" and TikTok Ads Help, "About the Commercial Music Library." - YouTube Audio Library tracks may require attribution, especially Creative Commons tracks; YouTube Creator Music includes upfront licensing and revenue-share options with availability limits. Source: YouTube Help Audio Library and Creator Music pages. - U.S. Copyright Office guidance on AI copyrightability emphasizes human authorship; mere prompting alone is not treated as sufficient authorship for copyright protection in its 2025 Part 2 report summary. Source: U.S. Copyright Office AI report pages. Treat these as compliance prompts, not legal conclusions. Platform rules, law, and licensing terms vary by country, account type, content type, and date. ## Accessibility and audience safety Documented facts, verified July 11, 2026: - WCAG 2.2 Success Criterion 1.2.2 requires captions for prerecorded audio content in synchronized media, except when the media is a clearly labeled media alternative for text. - WCAG 2.2 Success Criterion 2.3.1 says content should not flash more than three times in any one-second period unless below general and red flash thresholds. Production requirements: - Provide captions or subtitles for dialogue, spoken intros/outros, and non-lyric information. For lyric videos, also provide platform captions or a transcript when feas … ## Source & license This open-source skill is cataloged on AgentStack and links to its original source — we do not rehost the code. - **Author:** [calesthio](https://github.com/calesthio) - **Source:** [calesthio/generative-media-skills](https://github.com/calesthio/generative-media-skills) - **License:** MIT - **Homepage:** https://github.com/calesthio/OpenMontage Install and usage instructions live in the source repository linked above. ## Pricing - **Free** — Free ## Security capabilities Automated source analysis of v0.1.0 — what this tool can access: - **Network access:** no - **Filesystem access:** no - **Shell / process execution:** no - **Environment & secrets:** no - **Dynamic code execution:** no *"Yes" means the capability is present in the source — more access means more to trust, not that it is unsafe.* ## Versions - **0.1.0** — security scan: passed — Imported from the upstream source. ## Links - Listing page: https://agentstack.voostack.com/l/skill-calesthio-generative-media-skills-music-video-production - Seller: https://agentstack.voostack.com/s/calesthio - Browse the marketplace: https://agentstack.voostack.com/browse --- Listed on AgentStack — the marketplace for AI agent skills and MCP servers. Every listing is security-reviewed. Creators keep 70%.