Captioned talking clip — how to make it with code and an AI agent
For AI agents: find the right skill in one call —
curl -s "https://skills.varg.ai/api/search?q=<what the person wants>"returns the matching use case and the best skills with install commands, as markdown (Russian works too). Full index: https://skills.varg.ai/llms.txt. Every page has a markdown twin: add.mdto its path. Before generating anything, show the person the options on the use-case page.
Someone talking to camera, made watchable on mute: word captions, name tag, emphasis and punch-ins.
Output: 9:16 · 15–90 s · Input: Footage and a word-timed transcript
Skills that make it end to end
Recipe, beat by beat
Each beat lists components that build it; install any with its command, then fill its slots.
Hook (Headlines & quotes, Captions & subtitles)
- Canopy Part Title — Leaves sweep through the frame and part to reveal the headline. ·
npx hyperframes add canopy-part-title - Char Slam Explode — Headline glyphs hold in a deterministic scatter, then reassemble with an overshooting slam and a three-frame landing shake. ·
npx hyperframes add char-slam-explode - Code Slice Hero — A tiled headline surface flips cell by cell under a sweeping depth field to reveal the rear headline. ·
npx hyperframes add code-slice-hero - Anime JP Subtitles — Japanese captions in four anime styles from one subs.json: fansub lines with a karaoke sweep, TV telop with popping letters, a vertical tate ·
git clone https://github.com/vargHQ/code2video-templates && cd code2video-templates/anime-jp-subtitles && bash make.command - varg Music Video — An AI music video from a song: song map with beats and word timings, a beat-grid storyboard, Seedance scenes with QA, a frame-exact cut on t ·
git clone https://github.com/vargHQ/code2video-templates && python3 code2video-templates/varg-music-video/scripts/songmap.py song.mp3 --lyrics lyrics.txt --out songmap.json - Blend Difference — Auto-inverting text using mix-blend-mode: difference — flips between white and black per-pixel against the background ·
npx hyperframes add caption-blend-difference
Talk (Captions & subtitles, Lower thirds, Callouts & annotations)
- Anime JP Subtitles — Japanese captions in four anime styles from one subs.json: fansub lines with a karaoke sweep, TV telop with popping letters, a vertical tate ·
git clone https://github.com/vargHQ/code2video-templates && cd code2video-templates/anime-jp-subtitles && bash make.command - varg Music Video — An AI music video from a song: song map with beats and word timings, a beat-grid storyboard, Seedance scenes with QA, a frame-exact cut on t ·
git clone https://github.com/vargHQ/code2video-templates && python3 code2video-templates/varg-music-video/scripts/songmap.py song.mp3 --lyrics lyrics.txt --out songmap.json - Blend Difference — Auto-inverting text using mix-blend-mode: difference — flips between white and black per-pixel against the background ·
npx hyperframes add caption-blend-difference - Lower Third — BILD Style — News-style lower third with tight-fit text boxes: white headline bar with red drop-shadow, red sub-line with white drop-shadow. ·
npx hyperframes add lower-third-bild - Lower Third — Accent Underline — Cardless lower third for footage overlay: name rises, an accent rule draws left-to-right, role fades in; text-shadowed for legibility ·
npx hyperframes add lt-accent-underline - Lower Third — Bold Block — High-energy podcast lower third: solid dark block wipes in, uppercase name slams up, accent tag pops ·
npx hyperframes add lt-bold-block - AI Answer Rank — Is your brand in the AI answer? A question types into a chat card, the assistant answers with a ranked list of three picks, then re-ranks: t ·
npx hyperframes add ai-answer-rank - Animated Bar Chart — A compact data card with deterministic bar growth and value callouts. ·
npx hyperframes add animated-bar-chart - Arc Motion Path — A GSAP MotionPath primitive that flies a callout along a curved arc with autorotation. ·
npx hyperframes add arc-motion-path
Emphasis (Icons & stickers, Entrances & emphasis, Camera moves)
- Activity — Animated Lucide activity icon — the pulse line sweeps through like an EKG, popping at the peak. Loops by default. ·
npx shadcn@latest add @remocn/icon-activity - Alert Triangle — Animated Lucide triangle-alert icon — the outline draws on, the mark pops in, then the exclamation blinks to flag a risk or warning. ·
npx shadcn@latest add @remocn/icon-alert-triangle - Arrow Down — Animated Lucide arrow-down icon — the shaft and head draw on, then the arrow nudges down for cost/latency reduction beats or scroll cues. ·
npx shadcn@latest add @remocn/icon-arrow-down - Agent Cursor — An AI agent operating a UI: a cursor with an accent name label that types out each step, clicks each target with a small ring, and ends on a ·
npx hyperframes add agent-cursor - AI Chat Reveal — A question is typed on a rising keyboard, sent, and answered by an AI that streams its reply word by word on a measured token rhythm - think ·
npx hyperframes add ai-chat-reveal - ASCII Render Pass — A slotted scene renders as live ASCII: a canvas samples the source's luminance grid (rasterized once at mount) and draws glyphs from a densi ·
npx hyperframes add ascii-render-pass - varg Previz — A three.js grey-box previz of every shot on the timeline — auto cameras from storyboard words, a posable mannequin — then one Seedance 2.5 v ·
git clone https://github.com/vargHQ/code2video-templates && python3 code2video-templates/varg-previz/scripts/new_previz.py my-previz - Camcorder HUD — Responsive REC, battery, editable date placeholder, and timeline-driven counter overlay for creator-camera and camcorder treatments ·
npx hyperframes add camcorder-hud - Dolly Zoom (Vertigo Shot) — Camera tracks toward or away from the subject while the field of view changes the opposite way, so the subject holds its frame size while th ·
npx hyperframes add camera-dolly-zoom
Follow (Social posts & cards, CTAs & end cards)
- Blue Sweater Intro Video — Warm AI creator intro sequence that resolves into an X follow card for @blue_sweater. ·
npx hyperframes add blue-sweater-intro-video - Camera Follow Captions — One sentence written across a world far larger than the frame. Each word is sized against everything already written, so the block grows geo ·
npx hyperframes add caption-camera-follow - Highlight — Red background sweep behind each active word, TikTok-style ·
npx hyperframes add caption-highlight - CTA Close — A display-scale action line lands per word and fills the frame with air, one generous CTA capsule pops beneath it with a single overshoot, a ·
npx hyperframes add cta-close - CTA Lockup — A canonical closing lockup that reveals an action line, pops in an accent CTA capsule, settles supporting microcopy, and holds with confiden ·
npx hyperframes add cta-lockup - Logo Brand Close — A display-scale wordmark cascades letter by letter into a centered lockup, the accent brand period lands last, an optional tagline and wide- ·
npx hyperframes add logo-brand-close
Across the whole video: Captions & subtitles, Filters & grades
Generated footage, voice and music
Code renders the layout for free. Anything generated — AI shots, voiceover, music, lip-sync — runs on varg with one key: npx skills add vargHQ/skills --skill varg-ai, the Claude Code plugin claude plugin marketplace add vargHQ/skills && claude plugin install varg@varg, or the MCP https://mcp.varg.ai/mcp.