Video skills

Act as an AI Film Director, turn a rough idea, script, ad, or character concept into a render-ready cinematic production package for image+text AI video models (Veo3, Higgsfield, CoreReflex/WAN, and others).

Take a 9:16 talking-head video and overlay animated titles, step numbers, keyword badges, progress bars, and synced captions (Remotion/Hyperframe), original video stays full-frame.

Generate a 3-6s animated logo sting / brand intro or outro (Remotion/Hyperframe), logo reveal with spring scale, particle burst, tagline, brand-color glow.

Adapt an approved video master into platform-native aspect ratios, alternate hooks and frame-based covers while preserving approved identity and claims. Not a new campaign concept or text-only repurposing.

Direct and produce short cinematic ads from approved brand assets, visual references, or a new concept; optionally document the real production as a case study. Not for longform explainers, code-only motion graphics, or editing multiple camera takes.

Review an actual video against its approved brief/reference and revise only weak moments. Not code review or a promise of virality.

Turn a CSV (or pasted numbers) into an animated data-dashboard video (Remotion/Hyperframe), a hero KPI count-up, growing bar chart, self-drawing donut, and animated line chart.

Edit REAL recorded footage into a finished video, transcribe multi-take talking-head clips, pick the best take per scene, cut filler/warm-ups on word timing, stitch with FFmpeg, color grade, and add Remotion title/caption frames. The real-footage editing path (vs. as-video-hub which generates video from code).

Turn any topic into a fully animated explainer video (Remotion/Hyperframe), research, script, design, animate.

Produce a branded, character-led animated explainer end to end on the CoreReflex API, entity card intake, Format-N script, storyboard approval gate, one reused style token, illustrated plates + self-hosted VO + music bed, free engine retiming, verified render, and the measured case study. The character, palette and mission swap per entity; the pipeline is fixed.

Turn one topic or course module into a full multi-video explainer series, segmented outline, per-segment scripts, one shared design system, batch-rendered animated videos (via as-video-education-explainer-video), assembled into a playlist with thumbnails.

Produce a Format 1 Honest-Version explainer storyboard (7 chapters, 21 beats (HOOK / FACT / TURN / RULE), native-shape cards, sourced claims gate) as a self-contained HTML artifact and a CoreReflex longform composition brief. Not the 30-second Remotion explainer (as-video-education-explainer-video) and not cinematic AI video (as-video-ai-director).

Natural-language editing intent into deterministic FFmpeg pipelines (reframe to 9:16/1:1/16:9, burn captions, add a ducked music bed, trim, concat, transcode per platform, loudness-normalize, extract stills/GIFs) on existing footage, source preserved, output probe-verified.

The video front door (routes between EDITING real footage (as-video-edit-multitake) and GENERATING animated video from code (Remotion / Hyperframe)) explainers, product demos, testimonials, avatar overlays, data dashboards, social shorts, brand intros. Entry point for the video domain.

Shot bible and banned tropes for factory / test-stand / line film so industrial clients do not look like SaaS ads. Briefing is as-design-photoshoot-director; generation is as-video-ai-director.

Turn one video concept into an assembled long-form faceless video built from five-second blocks, each with a narration line, a standalone image prompt, and an image-to-video prompt. Sizes the block count from the target duration, writes a trailer-style thirty-second cold open, and keeps characters consistent across hundreds of prompts.

Render a real MP4 right inside Cowork using Pillow + ffmpeg. No Remotion, no GPU, no network. For fast, DISPOSABLE social stings, quote cards, and kinetic-text shorts when the Remotion pipeline can't run (e.g. the Cowork sandbox).

Turn any product URL into a full animated demo/launch video (Remotion/Hyperframe), scrape branding + real product images, rebuild the core UI, simulate a demo, show a showcase + CTA.

Run a productized AI-video service as a business, attract clients with a 4-part "revenue" video (credibility hook → real proof → method reveal → DM call-to-action), close in the DMs with a paid consult (no free calls), deliver with a 6-step staged-approval SOP (idea → script → cheap character render → produce → 90% cut with music → final).

Build a short-form video growth system for a filmed creator or operator, the profile as a conversion surface, viral modelling that studies breakouts and structure (not copying), retention-engineered production, and a weekly analyse-and-double-down loop over ninety days. Wins watch-time and completion first, reaction second, and converts the reach into followers and buyers.

Generate a punchy 10-20s vertical short (Remotion/Hyperframe) from a hook, tip, stat, or quote, pattern-interrupt open, kinetic text, burned-in captions, loop-friendly close.

End-to-end "drop a raw file, get a finished video" studio (video-use (transcript-first editing) cuts filler words, silences, retakes on word boundaries) then a word-synced motion-graphics engine of your choice, HyperFrames (HTML-native, default) or Remotion (React). It asks which.

Turn an image + a script into an AI talking-head (avatar) video for Reels/Shorts/TikTok/ads, generate the voiceover (TTS), drive a talking-person render (HeyGen/D-ID-style), then caption, add B-roll/overlays, and format vertical. For a presenter-style video without filming.

Turn a business's real Google reviews into an animated testimonial/social-proof video (Remotion/Hyperframe), scrape rating + reviews, animate stars + a review carousel + proof counters.

Convert any raw spoken script into TTS-ready narration that sounds human on any AI voice engine (the own-stack voice lane (CoreReflex cloned + stock voices) first, vendor engines (ElevenLabs-class) as fallback only) via strategic ALL-CAPS emphasis, dramatic ellipsis pauses, staccato rhythm, question lifts, and authority periods, while preserving every claim, statistic.