motion-shorts

Use when the user wants to create a vertical short-form video -- a 30-50s motion-graphics explainer with voiceover and captions for YouTube Shorts, TikTok, Instagram Reels, or LinkedIn. Handles the full pipeline from script to finished mp4. Defer to this skill whenever the user describes a video concept ("make a short about X", "generate a reel", "produce a video"), even if they don't mention Hyperframes or a specific format. Skip for very short teasers (under 15s) or when rendering an already-built episode.

Use when the user wants to start a new video episode -- scaffolding the project directory, starter scene-spec.json, and metadata under apps/hyperframe/src/episodes/. Defer to this skill whenever the user says "create a new episode", "scaffold a short", or "start a new video", even without mentioning the scaffolder tool directly. Skip for rendering or editing existing episodes.

Use when a Hyperframes short scene should use AI-generated raster images instead of pure HTML/CSS/SVG: product or workspace screenshots, handoff bundle diagrams, dense visual explainers, connector-heavy layouts, or any scene where card collisions, clipped text, or weak product-style visuals are likely. Produces prompt patterns, save-path/provenance rules, render-safe HTML usage, and frame-inspection checks for assets under apps/hyperframe/src/episodes/<slug>/assets/generated/.

Use when the user wants to produce a short from a source or idea, launched from this repo -- a blog URL, LinkedIn post, X/Twitter post, docs page ("hazme un short sobre <url>", "genera un short de este post"), or a freeform topic. Handles the end-to-end flow: capturing the source, drafting script alternatives with user approval, audio, scene-spec, per-scene QA, render, per-platform distribution copy, and archiving the result to the Notion Shorts Archive. The repo is the single entry point; Notion is only the downstream archive. Skip if the script already exists in the repo (use canonical-short directly).

Use when you need to prepare a presenter "talking head" cutout asset -- a transparent VP9-alpha video of a person speaking, plus its inverse-alpha plate -- for use in a short. Covers presenter sourcing (HeyGen or any MP4) and background removal via `hyperframes remove-background` -> `speaker.webm` + `plate.webm`. Defer to this skill when the user says "add a talking head", "presenter cutout", "avatar in the short", "HeyGen avatar", or asks to composite a person over motion-graphics. NOTE: the 17-type scene-hub has no avatar/cutout scene-type yet, so this skill currently produces the *asset* only -- scene integration is a documented follow-up (see "Integration status"). Skip for static portraits, static PFP overlays (use the `social-card` scene-type), or full-frame webcam footage with no alpha.

Use when the user needs to generate voiceover audio and word-level captions from a narration script -- the voice.mp3 + captions.json pair consumed by Hyperframes episodes. Handles ElevenLabs TTS + STT with cost guardrails and quality checks. Defer to this skill whenever the user mentions generating audio, TTS, voiceover, or captions for a video, even if they don't specify the pipeline or output format. Skip for transcribing pre-existing audio files (use npx hyperframes transcribe directly).

Convert text to speech using ElevenLabs voice AI. Use when generating audio from text, creating voiceovers, building voice apps, or synthesizing speech in 70+ languages.

Use when a Hyperframes short must be visually verified before claiming the work is done — the per-scene visual gate (Gate 3) of the production pipeline. Trigger for scene redesigns, overlap/clipping fixes, "looks bad" feedback, caption-safe checks, weak or crowded scenes, semantic mismatch, or any request that mentions screenshots, key frames, visual QA, malformed visuals, or overflow/overlap inspection.

Use when the user gives a text brief for a branded promo short -- "promo for the <event> webinar", "componme el promo de X con la marca Y". Composes scene-spec.json from the promo-* library (promo-intro-card, promo-hero, promo-card-speaker, promo-blur-cta, promo-agenda, promo-quote, promo-details), binds a brand pack via spec.brand, then assembles, runs per-scene QA, and renders. Defer to this skill for any brand-pack-driven promo. Skip for non-promo shorts (canonical-short) and when the brand pack does not exist yet (extract-brand-pack first).

Use when the user wants publishing copy for an episode -- YouTube Shorts title/description, TikTok caption, Instagram Reels caption, or a LinkedIn post matched to a rendered (or in-progress) short. Defer to this skill whenever the user says "write the copy", "dame la descripción", "prepare the LinkedIn post", or "copy de publicación" for an episode slug. Produces a validated src/episodes/<slug>/distribution.json with per-platform approval. Skip for drafting the narration script itself (that is canonical-short / produce-from-source).

Use when the user wants to capture a company's visual identity into a reusable brand pack -- "extract the brand from this Figma", "create a brand pack for <company>", "haz que los promos usen la marca X". Produces apps/hyperframe/brands/<slug>/brand.json plus composed logo SVGs and the brand font file so the promo-* scene-type library renders that brand. Defer to this skill whenever a new brand must flow through scene-spec "brand". Skip for one-off color tweaks on a single episode (use spec.palette) and for building new scene-types from a design reference.

Use when a Figma reference frame must become a new parametric scene-type under apps/hyperframe/templates/scenes/<type>/v1/ -- "replicate this frame as a scene-type", "add a scene like this design". Covers the 5-file contract, the bake-vs-HTML split, brand tokenisation, and pixel-measurement QA against the reference screenshot. Defer to this skill for any new scene-type sourced from a design reference. Skip when an existing scene-type can express the layout (fill slots instead) and for whole-episode composition (promo-from-brief).

Transcribe audio to text using ElevenLabs Scribe v2. Use when converting audio/video to text, generating subtitles, transcribing meetings, or processing spoken content.